Similar 75% price cut trajectory as GPT-5.6 Sol -> GPT-6.1 Sol. Overall, Terra is still half the price of Sol.
But like the original release of Terra, it's still a weird middle child between the small-model-that-got-substantially-better (Luna) and the large-model-that-got-significiantly-cheaper (Sol). The TerminalBench 4.0 cost-per-task results are a bit confusing and I'm not sure why you would pick Terra over Sol in that case. Overall this release seems like a flex on Sonnet 5.5: a model which is having its own middle child problems between the godly Opus 5.5 and the surprisingly capable Haiku 5.5.
Artificial Analysis still has MiMo 2.6 Pro setting the Pareto frontier where it matters: it's as cheap as Terra High/Sol Low while being smarter than Terra Xhigh. Sol Medium beats MiMo in smarts but is priced a lot higher. Meanwhile Terra Max is dominated by Sol High in both price-per-task and smarts.
It is having improvements in other fields than coding while being more cost efficient.
If all I want is coding at a cheaper price, then deepseek is a good model for that.
I am more than okay to have a bifurcation of models good for coding and models good for everything other than coding. I strangely imagined it many years ago.
Everyone is suggesting OpenAI to release a model close to opus 5.5 yet it's releasing this. On one hand we can imagine that maybe their training cycle wasn't able to produce this whereas gpt 6.1 terra was made.
But on the other hand, I also now wish to see an OpenAI model which is good for just pure coding, like an absolute beast at it while being cheap (so the opposite of terra)
ironically, one of the few appreciations that I might give to OpenAI if they pushed this model to prod and are facing some pushback as developers like us want/only see coding models. They could've benchmaxed it and it would've been slightly irrelevant. I'd much rather see reality.
At the very least, this new model won't make things more expensive for anybody and would only make use-cases cheaper depending on what they are, so it's a fine deal for me.
I also wonder maybe this is why open source labs aren't targeting this space and why the only focus of AI labs is coding nowadays. Everything else just doesn't seem so worth it from investor, even end user, training perspective.
Maybe this is the reason why we have gotten almost an exponential growth in coding capabilities as that became the most (if not only) capability that truly really matters.
To not wish for discussion to go to a tangent, nonetheless this could also explain why our work as a software engineer is going through a slight crisis of meaning as these AI models have gotten impressive at coding.
As an (aspiring) developer (currently in 1st year college), I don't wish for AI to replace coding even more for me or to push it even more or to suck the joy out of it even further. So I am more than happy for these models to stay as they are and rather make advancements elseplace. Ironic because it is us developers who use AI as well and we want cheaper workflows and so we would also want models to be cheaper and better at coding. There are some fine lines of contradictions and irony in all of it. I'd much rather also have open-weight models of such terra-like models but I imagine that because of the reasons that I mentioned above, we are most likely gonna see it the most focused on coding but still open-weights is pretty close.
The more choices the better (though choice-fatigue is a real thing), I hope I was able to explain some of the nuance I feel about this particular launch.
I work at OpenAI and I can try to explain our logic.
In general, LLM performance is incredibly multidimensional. There's no guarantee of a fixed ordering where the model that is best at math is also the model that's best at jokes. Despite that complexity, much of the variation in performance is explained by just a single variable: model size. In general, models that are bigger will (a) be more performant / reliable, (b) cost more per token, and (c) have better recall of esoteric facts. As you note, our models are ordered:
Luna < Terra < Sol < Astra
Even today, I'd recommend this as a general starting point for picking models. If you care about quality only, pick Astra. If you're a bit price sensitive, pick Sol. If you're very price sensitive, pick Terra or Luna.
Of course, model size doesn't capture all variation. Models that are trained later might have new data or new techniques or new architectures that earlier models lack. You can see this with models from all providers - Sonnet 5.5 outperforms Fable 5.1 in ways, and Gemini 3.8 Flash outperforms 3.1 Pro in ways.
In the case of GPT-6.1 Terra, I'd say Terra is probably better than Astra and Sol at technical communication, similar on vision and computer use (but much faster!), and a bit worse on most other things. For API developers, it's a nice midrange option if you want something cheaper than Sol but a lot smarter than Luna (especially for vision / computer use / communication). For subscribers, it's not a bombshell release - if you're hitting usage limits, maybe it helps you go a bit further, and if you want clearer communication or faster computer use, it can be a nice option to select.
If you have suggestions on how to name our products better, I'm all ears. Our intention is to make things clear and simple, even if we miss the mark sometimes.
Unfortunately, a lot of the complexity is somewhat irreducible and due to to the actual complexity of the products. Scaling model size is different from scaling reasoning effort which is different from scaling multi-agent count. So if we just swept all of those under a single name of "more IQ", it would make the naming simpler but the experience much worse, as important details would be obfuscated from end users.
Do you then recommend Sol for technical documentation?
What's confusing to me in addition to the naming is that there are no published guidelines I can use, as far as I can see. Put yourself in the user's shoes. On what basis should one pick a model for a given task?
I can understand smaller models being dumber and faster. That aligns well with scenarios where you are willing to compromise; e.g. you're classifying content or asking a model to format text into a table.
But when you want "maximum intelligence, cost and time be damned", there's the implication that bigger models are smarter on a linear scale, and every provider's model release naming is currently conflicting with that expectation.
This isn't just a naming problem, but a communication problem. If I knew that Astra was the best at coding but a bit bad at writing, and Sol was almost as smart but better at technical writing, then that's the kind of thing I expect to be documented.
One possible way: A simple matrix with horizontal columns like "writing", "coding", "math", rows are the models, and each box is a number of stars from 1-5.
Instead, this sort of thing becomes lore, shared by word of mouth.
> I'm all ears. Our intention to make things clear and simple, even if we miss the mark sometimes.
Thanks. I think the very response you wrote here is a better form of marketing for the software engineers than what openAI puts out there. When everything is groundbreaking and everything tops charts, nothing is.
I suspect a lot of target demographic (developers) prefer simpler and honest communication about not just the models high points but low points too. With the linked Terra post, if I saw 1 chart that clearly highlighted the usecases for which Terra is _worse_ than higher tier models, I would be more convinced overall.
Everything trying to occupy the top spot in my brain's limited "top model" representation space is not helpful.
The 4-tier naming is honestly as clear as you can get. I have no idea what people are so confused about.
Terra and Luna are dirt cheap and good enough for 50% (or more) of the typical junior dev's workload, and great for routine work. Sol and Astra have so much written about them at this point that it should be obvious what those are good at. The large/gigantic models for things that need full intelligence and slightly more branching thinking.
> If you have suggestions on how to name our products better, I'm all ears. Our intention is to make things clear and simple, even if we miss the mark sometimes.
I'm not the one you were talking with but catching you is godsend since I just ran into a lot of confusion on the naming myself. I think some of it has to do with how they are labeled in the app.
I think simple numbers in naming is great, even 6.1 is fine bc I understand that if I want the latest model, I just need to look for the biggest number. Where I got confused is in the "work" tab, I saw 'Astra' is set as the default. I see the label for it is:
"Frontier intelligence for your most demanding tasks."
To me that reads "this is the 'best' model." Then I see Sol-6.1 which is labeled:
"For coding and every day work."
Then Terra-6.1 is:
"Terra 6.1 Codex Model" (tbh this one might be the most confusing, I'm aware of codex the product but Sol is the one labeled good at coding)
From my POV as a basic plan-based user wanting to do something like...
> design a robust event-driven system, here's the details [...]
I honestly have no idea which one I should use.
Would Astra-6 produce a better design than Sol-6.1? Is the '.1' a big step up from Sol-6? (I'm guessing no but I truly have no idea). I would assume the answer is Astra-6 would produce the best one, even though the labels don't imply it is necessarily geared towards coding?
But tbh the "default" seems to always work for me... and after reading your comment, typing this one out, being forced to think about it... it doesn't seem too bad given your challenges in communicating the purposes of each model and wide audience of consumers. I'm also sure anyone could find a million problems with decisions I've made in software I've built so I try to not be too harsh :)
- Understanding how effort and class interact (Is a high effort terra above a low effort Sol? Are larger models mostly always better regardless of effort?)
- understanding when (if ever) a new model tends to be better than an older but higher class one. Is an old astra better than a new terra?
If increasing effort and increasing model class produce improved results in a different way, I miss making it explicit in which way. In other words: What kind of issue warrants going up in effort? what kind of issue warrants going up in model class? What kind of issue is unlikely to be solved by either?
> In general, LLM performance is incredibly multidimensional. There's no guarantee of a fixed ordering where the model that is best at math is also the model that's best at jokes.
Maybe stop using version numbers then and just name models like you'd name children.
Since internally you probably have a fairly good intuition for what these models do well, what about assembling a benchmark suite (with multiple dimensions) that produces results that show that intuition, but in a 'somewhat objective' way?
That could solve the 'more IQ' problem, because you'd actually have different dimensions (e.g. for 'raw intelligence', technical writing, emotional understanding, etc) that you could use to communicate the differences in models better.
(Of course people would instantly start complaining that the benchmarks numbers are probably fake etc, but it is what it is.)
Internal evals are still running but GPT-6.1 Terra is an AWESOME release for anyone using the API in agentic production applications where cost matters.
Personally, I pretty much exclusively use claude for coding work (sorry!). Our application is running almost 100% on OpenAI API (except for some fallback models). It is the best API on the market. The models are excellent particularly for any application that is cost sensitive.
Thank You for OP asking the question and thank you parents for some insight.
>If you have suggestions on how to name our products better......
Too late for that. I hope you guys can at least stick to these naming scheme for at least another 2-3 years going forward. Not everyone follows AI / LLM intensely, and it seems there is a new naming changes every year or so.
At least conceptually OpenAI's naming is easier to understand than Claude.
It terms of capability, it would be great to make it clear somewhere that this is 6.1 Model "+" technical communication, vision and computer use improvements. So that at least we know what we should try with these new release and choose a better model for certain workload.
Right now, other than pricing, it is confusing as hell. And in hindsight I wonder if there are fatigue in choosing one people just go with Open Model and stick with it if it is good enough.
What makes this even more confusing is reasoning effort.
You have Luna, Terra, Sol and Astra, each with 4+ reasoning levels. That's ~20 options to choose from.
On a repetitive task like customer support or something, where you have decent benchmarks, you can plot cost versus performance, but you can't really do that on coding.
I actually do have a recommendation, which is to do something like a new mode (i know i know):
Auto xlow
Auto low
Auto med
Auto high
Auto xhigh
Auto max
With an optional flash mode.
Then as you release models, you pick the model and thinking level that you believe is pareto optimal for cost/performance.
So this auto updates each release, like Terra was probably off this list entirely, but maybe this would appear now in Auto medium or something like that.
I know from looking at charts that each model seems to have sweet spots (like I always use this model in this thinking level and if i want to go cheaper i go here). If you picked those for us, it would be interesting. I do think you should keep the model families for those wanting more consistent experiences, but for many devs they may want to just pick auto something and have it update over time. I think nice if it actually just said in parenthesis what it was, something like:
Auto Medium (Terra 6.1 High)
I am not sure if everyone would agree here though. I'm obivously advocating for more naming schemes, and i understand the irony.
Please can I ask about one parameter you didn't mention here? All of your models have both a reasoning effort setting and also a non-pro vs pro setting. I have some intuitive feel (from experimenting) for the difference a larger vs smaller model makes, and why that's different from adjusting reasoning levels, but I have no real feel for how pro vs non-pro is different in behaviour from adjusting effort.
Appreciate that these things have orthogonal measurements; with that in mind, given that the name signifies size, what does version number signify?
I.e. are all version numbers independent and Model X 7.5 is not necessarily of the same generation or cut off as Model Y 7.5? Or do all of your 6.1 models have something in common - cut off data, generation, baseline technology, approach, architecture, or something else?
It’s probably just unavoidable. Like when Apple release a new MacBook Air first and it is (in some ways) faster than the old MacBook Pro.
I think the planets thing is a bit confusing though. I hadn’t really thought about it being based on the size of the planet. I assumed it was based on how far away they were with terra therefore being the base one.
I like to use OpenAI models for ingesting a customer brief, usually a paragraph or two and then a read of a several customer files (eg spreadsheets and docs etc) and turning it into a system design blueprint. I'm currently using Astra for this phase but it's incredibly expensive relative to the other tasks that come after this. Would you recommend I stick to Astra or can I downgrade to say Sol or Terra for this job?
I'm just worried the system design won't be as good if I choose a lesser model so to speak.
It would be interesting to know which model is internally viewed as being best at jokes, I have a personal humor benchmark and so far models will either steal jokes (from comedians like Maria Bamford), or write something very unfunny.
"In the case of GPT-6.1 Terra, I'd say Terra is probably better than Astra and Sol at technical communication" -- you're doing the thing here too, underspecifying your own models: 6.1 Terra is better than…which Sol, 6.1, 6, 5.6, just the latter two, all of them? This 2D bluriness is a big part of the confusion.
Straight up - why not just name them Cheapest, Cheaper, Expensive and More Expensive? Cost seems to be the main driver of the intuitive difference between models.
This kind of heuristic is actually extremely useful - and mostly what I try to develop over a couple weeks as the new models come out. I would love if we could get a fairly standardized take like this on the regular.
Of course we have far too many benchmarks (of questionable utility), but it would be helpful to get OpenAI's take on where it slots in, and how the landscape changes as these get rolled out.
> Models that are trained later might have new data or new techniques or new architectures that earlier models lack.
This seems like one of those cases, where CalVer would be appropriate, and you likely have ideas what you aim for while training, so technical model names (ids?) could include aliases to most things you said. Like
2026q4-xlarge-researcher, 2026q4-medium-humanitarian, 2026q4-medium-coder, 2027q1-small-analyst etc.
> important details would be obfuscated from end users.
I don't expect the vast majority of your users care. I would guess they want a rough proxy for "cost" and will assume cost=>capability are roughly linked in a linear fashion.
it is very possible that the users weighted by spend do, and I'm therefore simply wrong. But, 2c
> I work at OpenAI and I can try to explain our logic [...] general, models that are bigger will be more performant.
> If you have suggestions on how to name our products better, I'm all ears. Our intention is to make things clear and simple, even if we miss the mark sometimes.
How about naming the models:
"Small", "Medium", "Large" and "Very Large"
Or refer to them by parameter count like other companies (e.g. this is 6.1 5T)
If I have a clothing store, and rather than name the clothes sizes Small, Medium and Large I decide to call them Urgon, Planfar and Fangor I shouldn't be surprised if people think my size naming is confusing or silly. And IMO it's not a great explanation to say "oh it's easy, if you are skinny then you should get Urgon, but then if you are a bit more of a medium build you should get a Planfar, but if you want something a bit baggier than Fangor will really work out for you"
> If you have suggestions on how to name our products better, I'm all ears.
For some reason, this is just hilarious to me. Idk if you're being serious or if this is passive aggressive but the thought of OAI actually taking advice from a random HNer and basing their naming scheme off that is peak.
I don't understand what's so confusing about this.
The lineup of models is still the same.
There are four models: luna, terra, sol, and astra.
Pricing increases for each level. This new 6.1 Terra is 50% cheaper than Sol 6.1.
They are lowering prices for newer releases, so yes if you compare Sol 6.1 to 5.6 Terra it won't look right, but why are you doing that? Look at the same release version for comparable numbers.
Each model improves in successive releases but they maintain pricing and capability coherency within a release version.
Yes, Sol 6.1 is supposed to rival Astra for some work, which is what you'd expect from a later versioned release model against a prior release bigger model.
Assuming they release Astra 6.1 at some point, I expect it'll be better than Sol 6.1, at a higher price.
I don't understand what people are asking for here? No reductions in pricing? Different model names on every version release? Not releasing any models for a version number until they're all ready?
Hey! I think it's confusing because there is a mix of three components:
- Model name
- Version number
- Effort level
All 3 of these are a one-dimensional balance of price/speed, and capability (Loose terms). The confusion is in how these overlap, how they're announced/marketed etc. So you go from 1 dimensions to 3; you can visualize it as a box. For a given task, how do you choose where in this box you select? Which regions are your favorite? Can you map each region of this 3D space to specific tasks? Label them with intuitive names?
Given the operations of the model are opaque, can you see how someone (most people?) might be confused?
It's confusing because the models are released one at a time rather than as a family.
Let's say I have some task that Luna 6.1 cannot do, but Terra 6.1 can. So I build my product around that. In a month when Luna 6.2 is released, then I have to figure out whether it's worth to evaluate and swap or just wait for Terra 6.2.
I think a lot of confusion comes from the discombobulated rollout of these models as well.
They released gpt-6-sol (alongside astra and luna but no terra) and 7 days later released gpt-6.1-sol making gpt-6-sol basically DOA before any real adoption. Everyone knows it wasn’t just better RL/post-training, it was a completely different model underneath. Now they randomly reintroduce terra.
I’m sorry, but if you’re not even a tiny bit befuddled by this then you must be taking whatever OpenAI is having.
Part of the confusion is that it's not linear capabilities. One model is not strictly "better" than the other. They have orthogonal capabilities.
I think the names are more of a reference to price. Luna is always cheaper than terra which is cheaper than Sol which is cheaper than Astra (for the same model number).
But minor credit to Anthropic, at least they've mostly stuck to their four model names (Haiku, Sonnet, Opus, and Fable). You mostly know where a model is MEANT to be positioned from the name alone. OpenAI cannot keep it consistent from release to release.
But, yes, currently Opus Vs. Fable has the same issue as Sol-6.1 Vs. Astra.
It's like iPhone 17 being faster than iPhone 16 Pro. Because it's newer and benefits from advances in chip architecture, despite being a lower model tier. Sometimes different iPhone models are released at different times.
Both companies have screwed up their naming royally. Maybe that's intentional - I'm sure it doesn't hurt Anthropic to have people using Fable a bit longer than they need to and racking up a huge bill in the process.
If you daily all these models (and have a modest budget, not some >$400 corporate allowance), then it's pretty obvious that Luna is small and fast, and Sol 6.1 is the big one for jobs that need significantly above average intelligence, comparatively to a human engineer. I use both constantly. Astra is really not relevant for the average dev except to do some moonshot task with a vague verification loop.
Sure Sol 6.1 on low/medium can be considered "cheap" considering its abilities, but it's clearly not meant for the same type of tasks as Luna. You just grow an intuition for when something is solvable with a Luna-sized model. I'm guessing Terra is good for the tasks that you'd typically use Sol 6.1-low for, or for Luna type tasks (or other not-novel problem) where you want to reduce the risks of suboptimal decision making. It's exactly what I'd go with for feature additions where the correct solution is already mostly designed/specced-out.
It's like Intel i3/i5/i7/i9, or nvidia 50/60/70/80.
It is meant to give a rating inside a generation. Different generations are not necessarily comparable. an i7 is better than an i5, but not if you compare an i7 7xxx with an i5 14xxx.
A nvidia xx80 card is better than a xx60 card, but not if you compare a 1080 to a 5060.
The only difference for OpenAI is that not every generation has every model.
You're just not paying attention. If you run copilot in copilot using copilot effort, it performs much better than copilot in copilot using copilot effort. This is clearly explained if you visit copilot.com using copilot and then search for copilot.
The way to think of this is that (current) inferencing hardware has natural cut points of efficient scale (assuming some optimal architecture which at this point is almost always an MOE variant), so you always have an incentive to provide 3-4 model sizes (which nicely enough often means diff speed to answer also).
You are also rolling the training of your models by size and the lat est small model can sometimes beat an older model.
For a while we thought you just train the big model first then distill smaller models from it but since we've learnt the importance of on policy RL (basically the way a small model figures out a problem is very diff from a big model so it's best to RL optimize each model distinctly)
If you scale a model down, or switch out attention with linear attention, or introduce KV-compression, etc., you don't really know how it will affect model performance before you've trained the model.
And once you've spent millions training it, you might as well make it available if it's pareto optimal for some tasks. It also gives you more insight seeing how people use it in practice.
Agree, it's confusing. I hear people discussing which model to use for what tasks and my head starts to spin because there are so many properties to account for and it's generally hard to reason about LLMs.
"Speed, quality, price: pick two". I think I'd rather see a choice like that than the current convolution of $-per-token, sub quota, latency, simulated reasoning level, model size, context size, etc, etc.
>But then things go confusing, they released Sol-6.1 which sometimes is superior to Astra, and discontinued Terra which made it look like this
I don't want to rush to their defense here, but, when model families are getting refreshed every 6 weeks, there is going to be significant overlap, particularly with these staggered releases
Wait I always thought terra was the smaller model and the further away you went in space the bigger the model (luna is further away so it’s better than terra)
It's especially confusing because Sol 6.1 is mostly just Astra but 10x slower at 1/10th the price - unless you pay for it to be as fast as Astra - at which point it's the same speed and price...
It's nice that they have Terra which seems to fill a void that's been needed since Sol 6 focused on cost (at the expense of quality), Sol 6.1 got a huge boost in quality, but became unusably slow unless you paid Astra prices - at which point, why not just use Astra?
the only confusing bit was dropping terra for a bit,
Luna (Small), Terra (Medium), Sol (Big), Astra (Bigger)
is perfectly clear, sol beats astra sometimes because Astra 6.1 was scrapped. and note that opus 5.5 beats fable 5.1 most of the time too without a skipped model.
Different classes and different generations. Isn't it roughly the same as what car manufacturers do? Or even Macbooks? 2014 macbook pro has worse performance than 2026 macbook air, this doesn't confuse consumers.
They're doing its because they recently cut plan usage by about 50%, leaving Anthropic providing about 5x the value on their plans. [0]
The "cost per task" metric is just a weaselly way to sound like they didn't just double the cost for most users, and aren't having their proverbial lunch money stolen. They can no longer claim to provide more tokens, more capable models, or more usage in general. So they made up a new metric where "task" could be defined to mean whatever they need it to mean.
They had to become VERY specific in their wording to present any kind of value proposition that will entice people.
Cost per task, time per task, quality (good luck measuring objectively between models) are the metric that matter. I don't care how much a token costs, I'm not buying tokens, I'm buying solutions.
Interesting that both "none" and "low" reasoning levels appeared not to spend any reasoning tokens at all, and in fact low ended up using less tokens and costing less than none here.
"medium" did use 496 reasoning tokens, but that's apparently not enough here to return a visible reasoning trace.
It's neat that we can now compare the Terra without reasoning pelican (cost 0.64 cents) with the equivalently priced Luna with reasoning pelicans (Luna xhigh is 0.40 cents, Luna max is 0.91 cents). I think the Terra low one compares favorably to those, though I'd suggest also comparing them on tasks more representative of real work than generating pelican SVGs!
Looks great on the first chart, but terrible on the Terminal Bench 4.0 where it needs xhigh to come close to 6.1 Sol, and costs more for a slightly worse result.
Not necessarily, they may have selected benchmarks where Terra performs relatively well compared to those models. Pretty much every model release these days just cherry-picks & highlights the benchmarks where their latest model performs best. We'll have to wait for the complete benchmark suite to have a better picture.
OpenAI repeatedly keeps hurting user workflows by altering the model generations family for each new version in unpredictable ways. As such, it is not a sane ecosystem to develop in.
2019: We have no idea if computers will ever able to talk like a human in natural language.
2026: We have multiple magical know-everything oracles that can speak any language better than most humans and produce pages of output for fractions of a cent, but the model names are a bit confusing so let's just not bother.
> the model names are a bit confusing so let's just not bother.
I said neither of these two statements or claims. The fact is that OpenAI's model family and release is not standardized. This means that if I have applications that are using the high, mid, and low tier models, e.g. at v5.6, it's not straightforward for me to update to a new version, e.g. v6 or v6.1 of these models because there isn't immediately a 1:1 mapping at the time of release of the version. Competitors to OpenAI don't have this problem, so I will bother with them.
The implication is that 6.1 Terra should've been released together with 6.1 Sol and Luna on the same day, and that these tiers should stick across versions.
Why do Anthropic and OpenAI think it necessary to split their model into 4 versions (Opus, Sonnet, Fable, Haiku, Astra, Sol, Luna, Terra) vs. Google's 2 (Argon, Flash)?
Can anyone give me the compelling reason for Terra and Sonnet to exist?
Virtually every task I give to either is better served by going up a tier and down a reasoning level, or down a tier and up a reasoning level. I simply don't understand their use case. No testing I've done has shown them to have a niche where there isn't a cheaper-with-equal-results or more-performant-at-same-cost option.
I think for the majority of the audience here who use these models inside of coding agent harnesses w/ subsidized monthly subscriptions, there likely isn't one. Depending on how Terra contributes to the rate limits, it may make sense to use it for some subset of a repeated workflow to get more mileage out of sub.
If you pay API prices for inference in your product stack & have specific use-cases w/ custom evals, these smaller models provide an additional cost/speed lever to adjust. Ultimately, I think the point is to compete on cost for use-cases where Sol on the lowest reasoning is overkill.
If you have a fairly straightforward usecase where Terra performs just as well as Sol, but is faster and cheaper, why not use it?
Yes but you are missing the point. It's the case where Luna is not good enough, and Sol is too much. The problem is that on benchmarks Luna xhigh is typically comparable to Sol low in terms of capability. So where does Terra fit in?
On the API we use a lot of Luna (cost efficient) and sometimes Terra level models. I think we never use a larger model. These are models that our app runtimes depend on.
On subscription for coding and stuff I am basically on Astra Ultra all day long.
You should look at model reasoning level curves on the cost/performance chart on artificial analysis. If you compare the curves for haiku/sonnet/opus I would agree with you that sonnet serves no purpose because it is priced too high. If you look at luna/terra/sol you will see that terra is priced such that it successfully fills a gap between luna and sol. And when you add speed into the equation it adds another dimension where terra makes sense. terra low falls between luna xhigh and max on both cost and intelligence, but it is way faster because it uses way fewer reasoning tokens. So I would say there are workloads where terra makes sense. But from my analysis I don't see any workloads where sonnet makes sense.
There is a chasm of performance difference between haiku and sonnet ime. Sonnet is cheaper and faster than Opus, so if you're providing a service via api where cost and speed is important, it's a big difference.
Sonnet is way faster than Opus when doing everyday tasks, I think on the subscription plans it's a better option when you need something done faster than Opus' long thinking sessions.
I have some business automation process for some customers that I maintain, where Terra with low reasoning ended up with the best price/performance & latency. With an equivalent Luna with high reasoning, price was usually better, but quality had a higher variance (very much depends on where the thinking takes the model) and latency for high reasoning was worse.
If you use API in agent or your own integration with software, such models make a lot of sense. We use Terra 5.6 for "question to SQL" translation for example. It's good enough and affordable enough. Luna will probably not handle complex queries well while Sol is too expensive.
i use sonnet to generate answers to help-desk tickets, which it does for significantly cheaper and insignificantly worse than opus.
if you're just talking about use in codex or claude code, then sure, leave it on opus all the time. but for running routine api requests, sonnet is great.
I've been refused to do cybersecurity (read-only investigative access against the recent brotli cve) with Opus 5.5 when I had full SSH access, so I guess you've got to have a fallback?
I cannot figure it out either.
I use Astra Ultra to plan and 6.1Sol High or X-High to build and even on huge projects with 10s of 1000s of pages it is fast and competent
Cheap and affordable. Dispatching proper tasks to these models can save you a huge % of your quota. Especially if you _know_ the task is simple (ie add something here or there, or archive a bunch of files and reformat kinda thing)
Maybe it's my tasks, but in the cases you're describing, I'm usually better off going down to Luna. I specifically don't see the gap where "It's too complex for Luna-high" and "It's not complex enough for Sol-low" exists.
Just one thing about the advert, how can I be the only person here thinking the ad is weird - like for the marketing dep, its a bit weird the ad features the same tram that broke and tragically like ended some people - happened for like the first time ever in 100 years, like a year ago ( 14 months ago) just a bit weird.
Im honestly not trying to be negative, but its kinda weird I kinda feel I should mention it because noone else has?
Im sure literally anyone from southern europe must have thought this the second they saw the ad. It was big news last year.
Am I mad? Just a heads up for PR to always check things like this first when releasing a video? Maybe its just me? Just trying to help
the only things we've got to actually make a judgement on are vibes and token anxiety. vibes generally indicate that more expensive == better, but more expensive == more token anxiety. so you end up using the most expensive model x effort that keeps your anxiety in check. yeah, maybe terra is great at technical documents. i'm not going to use it because i'd need to spend a couple of days writing technical documents to work out if it was better than sol and then what if it isn't?
is opus 5.5 better than sol 6.1? hmm, i really can't tell. maybe claude is a bit more diligent and identifies more issues up front. but sol has this can do attitude and the codex harness just gets on and does it's job done.
The results for factual error rate on "difficult" prompts is staggering to me. Particularly the difference between 5.6-Terra and 6.1-Terra. These are wildly different Pareto fronts. Our custom enterprise agent is currently running on 5.6-Terra and it's already very good. I look forward to updating the configuration.
These "Pareto frontier" performance vs price graphs have become very popular and I find them mostly uninformative. Pricing is subjective. You chose the position on the X axis, and which measures to use on the Y Axis.
The error bars on eval graphs are commonly like ±5%. It's fairly common to get out-of-order points when the gap between them is only a few percentage points. If we ran it again, or ran it with minor alterations, I'd expect the ordering to be expected. In general, we don't massage any of our eval numbers and for good statistical hygiene we'll report the first run we do, even if later runs show slightly smoother curves.
They say it has "A significant step up on factuality and coding".
My browser is underlining "factuality" in red. AI hallucinations or hallucinatiality are still a big problem in my experience.
Unfortunately it suffers from the same problem as Luna: it will stop at an earliest excuse and no amount of prompting not to stop until X is done will convince it otherwise.
not gonna lie they did really bad and this drained my usage SO FAST. that model used lots of unnecessary skills and read through way more memory files needed.
If you are confused, that is how this company was always been, remember GPT 4, then they did some weird versioning.
I personally like Terra, it kind of does what I would ask Sol or Astra, but dont have to count tokens as much.
And Luna has been amazing tbh, it takes designs from prototype and in record time is implementing it, so that was a huge win.
But yeah, they always had confusing way of naming products and if you are not following news, you will be confused about what it is.
AA has limited 1017 benchmarks so some terras versus 1017 Max:
max - 3 points smarter, 50% more expensive
xhigh - same score, 25% cheaper
high - 1 point dumber, 50% cheaper
Yes, V4.1 Flash 1017 is my daily driver cloud model. Copes fine with what I'd consider difficult brownfield development, like fixing bugs in vLLM prompt cache and coming up with small/sane overlay patches for them. Maybe it's disappointing for people who want to one-shot a game or website from a single prompt; I haven't tried that.
It's nice that their peak-rate hours end quite early in the day UK time.
You can do real world development with weaker models than Deepseek Flash.
I still use SOTA for the main model though, but also I'm perfectly aware it's mostly out of FOMO.
Deepseek is more than enough for default subagents though. Any well defined and verifiable task and there's absolutely zero difference between SOTA and good-enough models. Used to be just speed, but Deepseek is fast.
That terminal bench 4.0 result is... bad? Terra XHigh more expensive than Sol Low and still does worse.
It's very likely this is a next checkpoint of GPT-6 "Sol", over-optimised for the web chat and offered to Codex subscribers as a "please go away" offering.
If this was half the price, I'd be very interested, but at 1/5/0.05? Feels like another panic release because they can't get Astra 6.1 out of the door...
People keep asking OpenAI for a mode that's as good as Opus 5.5, and they release this instead. Interesting strategy. Let's see how it works out for them.
They don't have anything competitive with Opus 5.5 yet. There is a supposed model in the works, "Bel" last time I paid attention to gossip, that is thought to be better, later this year. Allegedly Anthropic caught OpenAI off guard and / or their different training approaches ended up working better than anticipated.
My expectation is Fable 5.5 is next overall improvement, and my bet is OpenAI's "Bel" will be the next competitive one (or, their lack of safety culture will bite them/us in the process).
There’s also a chance they surrender the dev market altogether to Ant and just focus on the “normie” segment with dots, computer use, excel etc. In which case they won’t have anything competitive to opus-5.5 anytime soon for SWE.
This release is interesting for enterprise at API rates.
At least unless the Terminal Bench 4.0 score is more indicative than the other ones, because that one shows it clearly worse than 6.1 Sol for both performance and cost.
Taking the benchmarks (as reported by OpenAI) at face value, why wouldn't they show Opus 5.5 on the OS world benchmark, and show no Anthropic models at all on the "factual error rate" benchmark
Note that while that page is from Pangram, the studies on it are not. If they are excluding studies that show Pangram performs poorly in realistic conditions, please share those studies!
Similar 75% price cut trajectory as GPT-5.6 Sol -> GPT-6.1 Sol. Overall, Terra is still half the price of Sol.
But like the original release of Terra, it's still a weird middle child between the small-model-that-got-substantially-better (Luna) and the large-model-that-got-significiantly-cheaper (Sol). The TerminalBench 4.0 cost-per-task results are a bit confusing and I'm not sure why you would pick Terra over Sol in that case. Overall this release seems like a flex on Sonnet 5.5: a model which is having its own middle child problems between the godly Opus 5.5 and the surprisingly capable Haiku 5.5.
Artificial Analysis still has MiMo 2.6 Pro setting the Pareto frontier where it matters: it's as cheap as Terra High/Sol Low while being smarter than Terra Xhigh. Sol Medium beats MiMo in smarts but is priced a lot higher. Meanwhile Terra Max is dominated by Sol High in both price-per-task and smarts.
I just imagined reading this comment like 2.5 years ago :)
the surprisingly capable Haiku 5.5
Destroyed by Anthropics crazy decision to price it x5 after 100k context.
Packaging & pricing seem to be two things Anthropic really struggles with
do people really use luna post 100k? small models generally seems to be hallucinating hard for me once you are over that threshold.
I don't think software devs are the audience for 6.1 Terra, but instead computer work, automation, and technical writing.
Honestly I'd much rather take this model then.
It is having improvements in other fields than coding while being more cost efficient.
If all I want is coding at a cheaper price, then deepseek is a good model for that.
I am more than okay to have a bifurcation of models good for coding and models good for everything other than coding. I strangely imagined it many years ago.
Everyone is suggesting OpenAI to release a model close to opus 5.5 yet it's releasing this. On one hand we can imagine that maybe their training cycle wasn't able to produce this whereas gpt 6.1 terra was made.
But on the other hand, I also now wish to see an OpenAI model which is good for just pure coding, like an absolute beast at it while being cheap (so the opposite of terra)
ironically, one of the few appreciations that I might give to OpenAI if they pushed this model to prod and are facing some pushback as developers like us want/only see coding models. They could've benchmaxed it and it would've been slightly irrelevant. I'd much rather see reality.
At the very least, this new model won't make things more expensive for anybody and would only make use-cases cheaper depending on what they are, so it's a fine deal for me.
I also wonder maybe this is why open source labs aren't targeting this space and why the only focus of AI labs is coding nowadays. Everything else just doesn't seem so worth it from investor, even end user, training perspective.
Maybe this is the reason why we have gotten almost an exponential growth in coding capabilities as that became the most (if not only) capability that truly really matters.
To not wish for discussion to go to a tangent, nonetheless this could also explain why our work as a software engineer is going through a slight crisis of meaning as these AI models have gotten impressive at coding.
As an (aspiring) developer (currently in 1st year college), I don't wish for AI to replace coding even more for me or to push it even more or to suck the joy out of it even further. So I am more than happy for these models to stay as they are and rather make advancements elseplace. Ironic because it is us developers who use AI as well and we want cheaper workflows and so we would also want models to be cheaper and better at coding. There are some fine lines of contradictions and irony in all of it. I'd much rather also have open-weight models of such terra-like models but I imagine that because of the reasons that I mentioned above, we are most likely gonna see it the most focused on coding but still open-weights is pretty close.
The more choices the better (though choice-fatigue is a real thing), I hope I was able to explain some of the nuance I feel about this particular launch.
Well put.
I'm very confused about their model market position/naming schema.
First we had:
> Luna (Small), Terra (Medium), Sol (Big).
Which was fairly smooth after you got used to their planets.
But then they added Astra as the "Bigger," which was mostly still self-coherent:
> Luna (Small), Terra (Medium), Sol (Big), Astra (Bigger)
But then things go confusing, they released Sol-6.1 which sometimes is superior to Astra, and discontinued Terra which made it look like this:
> Luna (Small), Sol-6.1 == Astra
But with Sol-6.1 costing the old Terra-6.0 price, while, as I said performing as good as the old "Bigger" model in many non-3D/UI situations.
Now they're re-introduced Terra, as a 6.1 release, making this even more confusing.
Did Microsoft's chief branding officer move over to OpenAI?
I work at OpenAI and I can try to explain our logic.
In general, LLM performance is incredibly multidimensional. There's no guarantee of a fixed ordering where the model that is best at math is also the model that's best at jokes. Despite that complexity, much of the variation in performance is explained by just a single variable: model size. In general, models that are bigger will (a) be more performant / reliable, (b) cost more per token, and (c) have better recall of esoteric facts. As you note, our models are ordered:
Luna < Terra < Sol < Astra
Even today, I'd recommend this as a general starting point for picking models. If you care about quality only, pick Astra. If you're a bit price sensitive, pick Sol. If you're very price sensitive, pick Terra or Luna.
Of course, model size doesn't capture all variation. Models that are trained later might have new data or new techniques or new architectures that earlier models lack. You can see this with models from all providers - Sonnet 5.5 outperforms Fable 5.1 in ways, and Gemini 3.8 Flash outperforms 3.1 Pro in ways.
In the case of GPT-6.1 Terra, I'd say Terra is probably better than Astra and Sol at technical communication, similar on vision and computer use (but much faster!), and a bit worse on most other things. For API developers, it's a nice midrange option if you want something cheaper than Sol but a lot smarter than Luna (especially for vision / computer use / communication). For subscribers, it's not a bombshell release - if you're hitting usage limits, maybe it helps you go a bit further, and if you want clearer communication or faster computer use, it can be a nice option to select.
If you have suggestions on how to name our products better, I'm all ears. Our intention is to make things clear and simple, even if we miss the mark sometimes.
Unfortunately, a lot of the complexity is somewhat irreducible and due to to the actual complexity of the products. Scaling model size is different from scaling reasoning effort which is different from scaling multi-agent count. So if we just swept all of those under a single name of "more IQ", it would make the naming simpler but the experience much worse, as important details would be obfuscated from end users.
Do you then recommend Sol for technical documentation?
What's confusing to me in addition to the naming is that there are no published guidelines I can use, as far as I can see. Put yourself in the user's shoes. On what basis should one pick a model for a given task?
I can understand smaller models being dumber and faster. That aligns well with scenarios where you are willing to compromise; e.g. you're classifying content or asking a model to format text into a table.
But when you want "maximum intelligence, cost and time be damned", there's the implication that bigger models are smarter on a linear scale, and every provider's model release naming is currently conflicting with that expectation.
This isn't just a naming problem, but a communication problem. If I knew that Astra was the best at coding but a bit bad at writing, and Sol was almost as smart but better at technical writing, then that's the kind of thing I expect to be documented.
One possible way: A simple matrix with horizontal columns like "writing", "coding", "math", rows are the models, and each box is a number of stars from 1-5.
Instead, this sort of thing becomes lore, shared by word of mouth.
16 replies →
> I'm all ears. Our intention to make things clear and simple, even if we miss the mark sometimes.
Thanks. I think the very response you wrote here is a better form of marketing for the software engineers than what openAI puts out there. When everything is groundbreaking and everything tops charts, nothing is.
I suspect a lot of target demographic (developers) prefer simpler and honest communication about not just the models high points but low points too. With the linked Terra post, if I saw 1 chart that clearly highlighted the usecases for which Terra is _worse_ than higher tier models, I would be more convinced overall.
Everything trying to occupy the top spot in my brain's limited "top model" representation space is not helpful.
1 reply →
The 4-tier naming is honestly as clear as you can get. I have no idea what people are so confused about.
Terra and Luna are dirt cheap and good enough for 50% (or more) of the typical junior dev's workload, and great for routine work. Sol and Astra have so much written about them at this point that it should be obvious what those are good at. The large/gigantic models for things that need full intelligence and slightly more branching thinking.
32 replies →
> If you have suggestions on how to name our products better, I'm all ears. Our intention is to make things clear and simple, even if we miss the mark sometimes.
I'm not the one you were talking with but catching you is godsend since I just ran into a lot of confusion on the naming myself. I think some of it has to do with how they are labeled in the app.
I think simple numbers in naming is great, even 6.1 is fine bc I understand that if I want the latest model, I just need to look for the biggest number. Where I got confused is in the "work" tab, I saw 'Astra' is set as the default. I see the label for it is:
"Frontier intelligence for your most demanding tasks."
To me that reads "this is the 'best' model." Then I see Sol-6.1 which is labeled:
"For coding and every day work."
Then Terra-6.1 is:
"Terra 6.1 Codex Model" (tbh this one might be the most confusing, I'm aware of codex the product but Sol is the one labeled good at coding)
From my POV as a basic plan-based user wanting to do something like...
> design a robust event-driven system, here's the details [...]
I honestly have no idea which one I should use.
Would Astra-6 produce a better design than Sol-6.1? Is the '.1' a big step up from Sol-6? (I'm guessing no but I truly have no idea). I would assume the answer is Astra-6 would produce the best one, even though the labels don't imply it is necessarily geared towards coding?
But tbh the "default" seems to always work for me... and after reading your comment, typing this one out, being forced to think about it... it doesn't seem too bad given your challenges in communicating the purposes of each model and wide audience of consumers. I'm also sure anyone could find a million problems with decisions I've made in software I've built so I try to not be too harsh :)
I think the main problem is surfacing two topics:
- Understanding how effort and class interact (Is a high effort terra above a low effort Sol? Are larger models mostly always better regardless of effort?)
- understanding when (if ever) a new model tends to be better than an older but higher class one. Is an old astra better than a new terra?
If increasing effort and increasing model class produce improved results in a different way, I miss making it explicit in which way. In other words: What kind of issue warrants going up in effort? what kind of issue warrants going up in model class? What kind of issue is unlikely to be solved by either?
6 replies →
> In general, LLM performance is incredibly multidimensional. There's no guarantee of a fixed ordering where the model that is best at math is also the model that's best at jokes.
Maybe stop using version numbers then and just name models like you'd name children.
Since internally you probably have a fairly good intuition for what these models do well, what about assembling a benchmark suite (with multiple dimensions) that produces results that show that intuition, but in a 'somewhat objective' way?
That could solve the 'more IQ' problem, because you'd actually have different dimensions (e.g. for 'raw intelligence', technical writing, emotional understanding, etc) that you could use to communicate the differences in models better.
(Of course people would instantly start complaining that the benchmarks numbers are probably fake etc, but it is what it is.)
2 replies →
Internal evals are still running but GPT-6.1 Terra is an AWESOME release for anyone using the API in agentic production applications where cost matters.
Personally, I pretty much exclusively use claude for coding work (sorry!). Our application is running almost 100% on OpenAI API (except for some fallback models). It is the best API on the market. The models are excellent particularly for any application that is cost sensitive.
It's not gonna happen but:
> Despite that complexity, much of the variation in performance is explained by just a single variable: model size.
please release model size :p
I have gone deep in dealing with naming and names.
Versions are good for backend but when communicating I want 2 or 3 levels of detail.
Class_Generation_Modifier
Astra_8_Fast
That's it.
This means: Class - Tells me what it can do, how well. Generation X end of life is ~18 months. Modifier: Variation of options.
And the modifier can be dropped if you add more classes.
Thank You for OP asking the question and thank you parents for some insight.
>If you have suggestions on how to name our products better......
Too late for that. I hope you guys can at least stick to these naming scheme for at least another 2-3 years going forward. Not everyone follows AI / LLM intensely, and it seems there is a new naming changes every year or so.
At least conceptually OpenAI's naming is easier to understand than Claude.
It terms of capability, it would be great to make it clear somewhere that this is 6.1 Model "+" technical communication, vision and computer use improvements. So that at least we know what we should try with these new release and choose a better model for certain workload.
Right now, other than pricing, it is confusing as hell. And in hindsight I wonder if there are fatigue in choosing one people just go with Open Model and stick with it if it is good enough.
This was lovely. Can we have more of this style of communication?
Instead of benchkmarks it'd be useful to get a sort of Radar chart across measured capabilities like computer use, communication, coding, etc.
We're probably headed to further specialized agents where a single ordering or even model size won't mean a lot.
What makes this even more confusing is reasoning effort.
You have Luna, Terra, Sol and Astra, each with 4+ reasoning levels. That's ~20 options to choose from.
On a repetitive task like customer support or something, where you have decent benchmarks, you can plot cost versus performance, but you can't really do that on coding.
I actually do have a recommendation, which is to do something like a new mode (i know i know): Auto xlow Auto low Auto med Auto high Auto xhigh Auto max
With an optional flash mode.
Then as you release models, you pick the model and thinking level that you believe is pareto optimal for cost/performance.
So this auto updates each release, like Terra was probably off this list entirely, but maybe this would appear now in Auto medium or something like that.
I know from looking at charts that each model seems to have sweet spots (like I always use this model in this thinking level and if i want to go cheaper i go here). If you picked those for us, it would be interesting. I do think you should keep the model families for those wanting more consistent experiences, but for many devs they may want to just pick auto something and have it update over time. I think nice if it actually just said in parenthesis what it was, something like: Auto Medium (Terra 6.1 High)
I am not sure if everyone would agree here though. I'm obivously advocating for more naming schemes, and i understand the irony.
1 reply →
Please can I ask about one parameter you didn't mention here? All of your models have both a reasoning effort setting and also a non-pro vs pro setting. I have some intuitive feel (from experimenting) for the difference a larger vs smaller model makes, and why that's different from adjusting reasoning levels, but I have no real feel for how pro vs non-pro is different in behaviour from adjusting effort.
3 replies →
Appreciate that these things have orthogonal measurements; with that in mind, given that the name signifies size, what does version number signify?
I.e. are all version numbers independent and Model X 7.5 is not necessarily of the same generation or cut off as Model Y 7.5? Or do all of your 6.1 models have something in common - cut off data, generation, baseline technology, approach, architecture, or something else?
Thanks! :)
1 reply →
>> If you have suggestions on how to name our products better, I'm all ears
<version>-<size>
Where <version>: 6.0, 6.1, etc., and <size>: xs | sm | md | lg
E.g. 6.1 Terra = 6.1-sm
If you eventually make specializations like "math" and "jokes" then I would call it something like 6.1-jokes-sm.
This makes a lot of sense.
It’s probably just unavoidable. Like when Apple release a new MacBook Air first and it is (in some ways) faster than the old MacBook Pro.
I think the planets thing is a bit confusing though. I hadn’t really thought about it being based on the size of the planet. I assumed it was based on how far away they were with terra therefore being the base one.
I like to use OpenAI models for ingesting a customer brief, usually a paragraph or two and then a read of a several customer files (eg spreadsheets and docs etc) and turning it into a system design blueprint. I'm currently using Astra for this phase but it's incredibly expensive relative to the other tasks that come after this. Would you recommend I stick to Astra or can I downgrade to say Sol or Terra for this job?
I'm just worried the system design won't be as good if I choose a lesser model so to speak.
Edit: typos/grammar
It would be interesting to know which model is internally viewed as being best at jokes, I have a personal humor benchmark and so far models will either steal jokes (from comedians like Maria Bamford), or write something very unfunny.
1 reply →
"In the case of GPT-6.1 Terra, I'd say Terra is probably better than Astra and Sol at technical communication" -- you're doing the thing here too, underspecifying your own models: 6.1 Terra is better than…which Sol, 6.1, 6, 5.6, just the latter two, all of them? This 2D bluriness is a big part of the confusion.
1 reply →
Straight up - why not just name them Cheapest, Cheaper, Expensive and More Expensive? Cost seems to be the main driver of the intuitive difference between models.
3 replies →
This kind of heuristic is actually extremely useful - and mostly what I try to develop over a couple weeks as the new models come out. I would love if we could get a fairly standardized take like this on the regular.
Of course we have far too many benchmarks (of questionable utility), but it would be helpful to get OpenAI's take on where it slots in, and how the landscape changes as these get rolled out.
> Models that are trained later might have new data or new techniques or new architectures that earlier models lack.
This seems like one of those cases, where CalVer would be appropriate, and you likely have ideas what you aim for while training, so technical model names (ids?) could include aliases to most things you said. Like 2026q4-xlarge-researcher, 2026q4-medium-humanitarian, 2026q4-medium-coder, 2027q1-small-analyst etc.
> important details would be obfuscated from end users.
I don't expect the vast majority of your users care. I would guess they want a rough proxy for "cost" and will assume cost=>capability are roughly linked in a linear fashion.
it is very possible that the users weighted by spend do, and I'm therefore simply wrong. But, 2c
In your own response you mentioned GPT-6.1 Terra, but I've seen GPT Terra-6.1. I would start by the part that is the most powerful:
GPT 6.1 X > GPT 6.0 X GPT Y Astra > GPT Y Sol > GPT Y Terra > GPT Y Luna
This order is very important.
Use colloquial terms: Big, Fast, Genius, Affordable, etc.
2 replies →
What prevents OAI (and Ant for that matter) from releasing new version for all models at once?
Instead, we got Opus 5.5, Sonnet 5.5 and Haiku 5.5 released weeks apart, and so are Sol 6.1 and Terra 6.1?
3 replies →
> I work at OpenAI and I can try to explain our logic [...] general, models that are bigger will be more performant.
> If you have suggestions on how to name our products better, I'm all ears. Our intention is to make things clear and simple, even if we miss the mark sometimes.
How about naming the models:
"Small", "Medium", "Large" and "Very Large"
Or refer to them by parameter count like other companies (e.g. this is 6.1 5T)
If I have a clothing store, and rather than name the clothes sizes Small, Medium and Large I decide to call them Urgon, Planfar and Fangor I shouldn't be surprised if people think my size naming is confusing or silly. And IMO it's not a great explanation to say "oh it's easy, if you are skinny then you should get Urgon, but then if you are a bit more of a medium build you should get a Planfar, but if you want something a bit baggier than Fangor will really work out for you"
3 replies →
The length of your explanation shows just how bad of a naming scheme this is.
> Our intention is to make things clear and simple, even if we miss the mark sometimes.
Well you definitely missed that mark extremely badly here.
3 replies →
> If you have suggestions on how to name our products better, I'm all ears
Terra 2026.2
Is there someone I can email directly about fine-tuning - all my models have been queued for 24/48 hours and your customer support seems non-existent
Fwiw, if you have to explain your product's name to the customers who just want to use it, that's a sign the products are confusing.
Tiny/2026-11. Medium/2026-6. Big/2026-10. XL/2026-01. It may not be sexy, but people know exactly what it is and what to use.
If there's no fixed ordering, can we have model names that describe the capability?
Have a model alias called "math", another "jokes". Quality labels/date suffixes to these obvious labels. Repoint as needed.
Something for people who don't want to play the cute "which part of the solar system is trending this week" game.
2 replies →
What do you understand by “technical communication”?
Would you like to recommend a model that does not produce errors and can be relied on like a traditional computer, please.
You certainly know that it happened not by accident that traditional computers are producing reliable results, a lot of work went into this.
Computers must produce reliable results, they are simply not useful if not.
People will understand that at some point, but a lot of damage already has been done.
1 reply →
naming suggestion: small, medium, large, x-large
it's almost like ya'll let the AI name it ;)
3 replies →
> If you have suggestions on how to name our products better, I'm all ears.
Hire an actual marketing person instead of letting engineers / ML researchers do things they're clearly terrible at ?
[dead]
[dead]
[flagged]
[flagged]
> If you have suggestions on how to name our products better, I'm all ears.
For some reason, this is just hilarious to me. Idk if you're being serious or if this is passive aggressive but the thought of OAI actually taking advice from a random HNer and basing their naming scheme off that is peak.
4 replies →
I don't understand what's so confusing about this.
The lineup of models is still the same.
There are four models: luna, terra, sol, and astra.
Pricing increases for each level. This new 6.1 Terra is 50% cheaper than Sol 6.1.
They are lowering prices for newer releases, so yes if you compare Sol 6.1 to 5.6 Terra it won't look right, but why are you doing that? Look at the same release version for comparable numbers.
Each model improves in successive releases but they maintain pricing and capability coherency within a release version.
Yes, Sol 6.1 is supposed to rival Astra for some work, which is what you'd expect from a later versioned release model against a prior release bigger model.
Assuming they release Astra 6.1 at some point, I expect it'll be better than Sol 6.1, at a higher price.
I don't understand what people are asking for here? No reductions in pricing? Different model names on every version release? Not releasing any models for a version number until they're all ready?
Hey! I think it's confusing because there is a mix of three components:
All 3 of these are a one-dimensional balance of price/speed, and capability (Loose terms). The confusion is in how these overlap, how they're announced/marketed etc. So you go from 1 dimensions to 3; you can visualize it as a box. For a given task, how do you choose where in this box you select? Which regions are your favorite? Can you map each region of this 3D space to specific tasks? Label them with intuitive names?
Given the operations of the model are opaque, can you see how someone (most people?) might be confused?
10 replies →
It's confusing because the models are released one at a time rather than as a family.
Let's say I have some task that Luna 6.1 cannot do, but Terra 6.1 can. So I build my product around that. In a month when Luna 6.2 is released, then I have to figure out whether it's worth to evaluate and swap or just wait for Terra 6.2.
I think a lot of confusion comes from the discombobulated rollout of these models as well.
They released gpt-6-sol (alongside astra and luna but no terra) and 7 days later released gpt-6.1-sol making gpt-6-sol basically DOA before any real adoption. Everyone knows it wasn’t just better RL/post-training, it was a completely different model underneath. Now they randomly reintroduce terra.
I’m sorry, but if you’re not even a tiny bit befuddled by this then you must be taking whatever OpenAI is having.
Part of the confusion is that it's not linear capabilities. One model is not strictly "better" than the other. They have orthogonal capabilities.
I think the names are more of a reference to price. Luna is always cheaper than terra which is cheaper than Sol which is cheaper than Astra (for the same model number).
> I think the names are more of a reference to price. Luna is always cheaper than terra which is cheaper than Sol which is cheaper than Astra
Per token, yes. But per task it's not even necessarily the case depending on how much thinking is needed.
1 reply →
I guess it's the same as Opus 5.5 being in a way better than any Fable.
That's definitely a problem too.
But minor credit to Anthropic, at least they've mostly stuck to their four model names (Haiku, Sonnet, Opus, and Fable). You mostly know where a model is MEANT to be positioned from the name alone. OpenAI cannot keep it consistent from release to release.
But, yes, currently Opus Vs. Fable has the same issue as Sol-6.1 Vs. Astra.
1 reply →
I don't know why this is so hard to understand.
It's like iPhone 17 being faster than iPhone 16 Pro. Because it's newer and benefits from advances in chip architecture, despite being a lower model tier. Sometimes different iPhone models are released at different times.
Yeah this isn't intuitive at all.
Both companies have screwed up their naming royally. Maybe that's intentional - I'm sure it doesn't hurt Anthropic to have people using Fable a bit longer than they need to and racking up a huge bill in the process.
4 replies →
If you daily all these models (and have a modest budget, not some >$400 corporate allowance), then it's pretty obvious that Luna is small and fast, and Sol 6.1 is the big one for jobs that need significantly above average intelligence, comparatively to a human engineer. I use both constantly. Astra is really not relevant for the average dev except to do some moonshot task with a vague verification loop.
Sure Sol 6.1 on low/medium can be considered "cheap" considering its abilities, but it's clearly not meant for the same type of tasks as Luna. You just grow an intuition for when something is solvable with a Luna-sized model. I'm guessing Terra is good for the tasks that you'd typically use Sol 6.1-low for, or for Luna type tasks (or other not-novel problem) where you want to reduce the risks of suboptimal decision making. It's exactly what I'd go with for feature additions where the correct solution is already mostly designed/specced-out.
It's like Intel i3/i5/i7/i9, or nvidia 50/60/70/80.
It is meant to give a rating inside a generation. Different generations are not necessarily comparable. an i7 is better than an i5, but not if you compare an i7 7xxx with an i5 14xxx.
A nvidia xx80 card is better than a xx60 card, but not if you compare a 1080 to a 5060.
The only difference for OpenAI is that not every generation has every model.
> Did Microsoft's chief branding officer move over to OpenAI?
Obviously not, otherwise the four models would be called Copilot, as well as the harness.
You're just not paying attention. If you run copilot in copilot using copilot effort, it performs much better than copilot in copilot using copilot effort. This is clearly explained if you visit copilot.com using copilot and then search for copilot.
“Introducing our first mobile devices: Pocket Copilot Luna for on-the-go inference, and our new tablets, Copilot Sol+ and Copilot Astra Duo.”
The way to think of this is that (current) inferencing hardware has natural cut points of efficient scale (assuming some optimal architecture which at this point is almost always an MOE variant), so you always have an incentive to provide 3-4 model sizes (which nicely enough often means diff speed to answer also).
You are also rolling the training of your models by size and the lat est small model can sometimes beat an older model.
For a while we thought you just train the big model first then distill smaller models from it but since we've learnt the importance of on policy RL (basically the way a small model figures out a problem is very diff from a big model so it's best to RL optimize each model distinctly)
Name means size which means price
Yes, a more modern smaller model can beat an older bigger model at some tasks, that's what can be confusing.
Training models is still not a precise science.
If you scale a model down, or switch out attention with linear attention, or introduce KV-compression, etc., you don't really know how it will affect model performance before you've trained the model.
And once you've spent millions training it, you might as well make it available if it's pareto optimal for some tasks. It also gives you more insight seeing how people use it in practice.
Flip it around. The number (generation) is the most important. Just like it's often with hardware.
And some variants (e.g., the medium one) being skipped or delayed for a certain generation doesn't really change the overall size branding.
Agree, it's confusing. I hear people discussing which model to use for what tasks and my head starts to spin because there are so many properties to account for and it's generally hard to reason about LLMs.
"Speed, quality, price: pick two". I think I'd rather see a choice like that than the current convolution of $-per-token, sub quota, latency, simulated reasoning level, model size, context size, etc, etc.
>But then things go confusing, they released Sol-6.1 which sometimes is superior to Astra, and discontinued Terra which made it look like this
I don't want to rush to their defense here, but, when model families are getting refreshed every 6 weeks, there is going to be significant overlap, particularly with these staggered releases
And low/medium/high/xhigh/ultra reasoning. And pro mode and ultra-speed versions. There's at least 25 options to choose from.
Next AGI test, when model companies have consistently good version branding.
Wait I always thought terra was the smaller model and the further away you went in space the bigger the model (luna is further away so it’s better than terra)
The moon is smaller than the earth. The earth is smaller than the sun. The sun is smaller than all of outer space.
It's especially confusing because Sol 6.1 is mostly just Astra but 10x slower at 1/10th the price - unless you pay for it to be as fast as Astra - at which point it's the same speed and price...
It's nice that they have Terra which seems to fill a void that's been needed since Sol 6 focused on cost (at the expense of quality), Sol 6.1 got a huge boost in quality, but became unusably slow unless you paid Astra prices - at which point, why not just use Astra?
Now at least we have a viable other option again.
Huh? I have not observed Sol 6.1 and Astra 6 to be the same.
You're not alone on this. They have too many models, and too many effort levels, it starts to feel like what to chose.
the only confusing bit was dropping terra for a bit,
Luna (Small), Terra (Medium), Sol (Big), Astra (Bigger)
is perfectly clear, sol beats astra sometimes because Astra 6.1 was scrapped. and note that opus 5.5 beats fable 5.1 most of the time too without a skipped model.
No idea, i still try to figure out how to use the right one per task.
Different classes and different generations. Isn't it roughly the same as what car manufacturers do? Or even Macbooks? 2014 macbook pro has worse performance than 2026 macbook air, this doesn't confuse consumers.
I suspect 6.1 Astra is the model that kept showing up in the news stories about an internal unreleased model trying to hack the Pentagon or whatever.
also am I the only person who remembers Do Kwon and the Terra/Luna debacle?? how about Gaia/Selene (admittedly harder to non-native english speakers)
[dead]
[dead]
Bet on purpose
“Test them all”
That's how their pricing tiers end up now. https://openai.com/business/pricing/#api
(For some reason they say "Cache read" instead of "Cached input" for 6.1 Terra)
6.1 Astra and 6.1 Luna must be coming up soon. Will they reduce their prices further?
I guess 5x/95% is where we're landing with the 6.1 family, that's nice.
[dead]
I rather enjoy OpenAI's focus on cost-per-task.
I'm optimistic about them tackling frontier intelligence next. Astra 6.1 please?
They're doing its because they recently cut plan usage by about 50%, leaving Anthropic providing about 5x the value on their plans. [0]
The "cost per task" metric is just a weaselly way to sound like they didn't just double the cost for most users, and aren't having their proverbial lunch money stolen. They can no longer claim to provide more tokens, more capable models, or more usage in general. So they made up a new metric where "task" could be defined to mean whatever they need it to mean.
They had to become VERY specific in their wording to present any kind of value proposition that will entice people.
[0] https://newsletter.semianalysis.com/p/anthropic-subscription...
> leaving Anthropic providing about 5x the value on their plans
5x is calculated in terms of API costs, nobody pays those extreme Anthropic prices outside enterprise users.
> like they didn't just double the cost for most users
Most users aren't on their $200 plan. Most users (including on the $200 plan) continue being subsidized.
4 replies →
Cost per task, time per task, quality (good luck measuring objectively between models) are the metric that matter. I don't care how much a token costs, I'm not buying tokens, I'm buying solutions.
4 replies →
I kind of expect them to go right to Astra 6.2, 6.1 was cancelled over the cybersecurity hubbub. I eagerly await either one.
Pelicans for GPT-6.1 Terra: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
Interesting that both "none" and "low" reasoning levels appeared not to spend any reasoning tokens at all, and in fact low ended up using less tokens and costing less than none here.
"medium" did use 496 reasoning tokens, but that's apparently not enough here to return a visible reasoning trace.
OpenAI's current frontier model lineup is gpt-6-astra, gpt-6.1-sol, gpt-6.1-terra, gpt-6-luna, so here's a comparison grid of all of their pelicans at different reasoning levels: https://static.simonwillison.net/static/2026/openai-pelicans...
What's the point of using it without reasoning? Isn't that what Luna's for?
It's neat that we can now compare the Terra without reasoning pelican (cost 0.64 cents) with the equivalently priced Luna with reasoning pelicans (Luna xhigh is 0.40 cents, Luna max is 0.91 cents). I think the Terra low one compares favorably to those, though I'd suggest also comparing them on tasks more representative of real work than generating pelican SVGs!
a bigger model still has more built-in knowledge to draw from, even without reasoning.
I appreciate that most of these pelicans are not holding the handlebars, i think that's more realistic than the alternative
I can literally see what the training data sample looked like that they included, so it performs well on this benchmark.
12 seconds for a pelican on low. Pretty good.
Do you have an agent scanning the front page waiting to post the pelicans?
I should do but no, it's just a bad HN habit.
Seems Gemini is way ahead in pelican drawing
Apparently Google made it a point to train their models to output SVG, among other things, so that helps
I don't understand, in my tests, on max effort, it's worse than 6.1 sol and 2x more expensive[0].
Used on medium seems to be the best use-case[1] and cheaper than 6.1 Sol.
[0]: https://aibenchy.com/compare/openai-gpt-6-1-sol-max/openai-g...
[1]: https://aibenchy.com/compare/openai-gpt-6-1-sol-medium/opena...
It is fast, but more verbose (a lot more verbose than Sol at higher efforts)
PacBench results.
GPT-6.1 Terra scored 84, one point over Grok 4.7 (83), in 5.4 min for $0.19 vs Grok’s 36.5 min and $1.69. The cheap models keep getting better.
GPT-6.1 Terra: https://jonclegg.github.io/pacman-bakeoff/#gpt-6.1-terra
Closest competitor (Grok 4.7): https://jonclegg.github.io/pacman-bakeoff/#grok-4.7
Does it suitable for finance choice?
Too many models and too many knobs. There are like 5 models and 5 thinking levels, that is 25 different combinations.
I agree but innovation does involve jabbing in many directions. They're trying to innovate.
Wow. If those charts are true, this effectively matches GPT-6-Astra at a fraction of the cost, and is neck-and-neck with 6.1-Sol.
Looks great on the first chart, but terrible on the Terminal Bench 4.0 where it needs xhigh to come close to 6.1 Sol, and costs more for a slightly worse result.
I would guess only on the nitpicked benchmarks?
The pelicans place it nicely between Sol and Luna in terms of both pricing and results.
pelican is also nitpicked (and probably the most useless)
3 replies →
Not necessarily, they may have selected benchmarks where Terra performs relatively well compared to those models. Pretty much every model release these days just cherry-picks & highlights the benchmarks where their latest model performs best. We'll have to wait for the complete benchmark suite to have a better picture.
OpenAI repeatedly keeps hurting user workflows by altering the model generations family for each new version in unpredictable ways. As such, it is not a sane ecosystem to develop in.
2019: We have no idea if computers will ever able to talk like a human in natural language. 2026: We have multiple magical know-everything oracles that can speak any language better than most humans and produce pages of output for fractions of a cent, but the model names are a bit confusing so let's just not bother.
> the model names are a bit confusing so let's just not bother.
I said neither of these two statements or claims. The fact is that OpenAI's model family and release is not standardized. This means that if I have applications that are using the high, mid, and low tier models, e.g. at v5.6, it's not straightforward for me to update to a new version, e.g. v6 or v6.1 of these models because there isn't immediately a 1:1 mapping at the time of release of the version. Competitors to OpenAI don't have this problem, so I will bother with them.
What is your implication here, in the case of 6.1 Terra?
The implication is that 6.1 Terra should've been released together with 6.1 Sol and Luna on the same day, and that these tiers should stick across versions.
Why do Anthropic and OpenAI think it necessary to split their model into 4 versions (Opus, Sonnet, Fable, Haiku, Astra, Sol, Luna, Terra) vs. Google's 2 (Argon, Flash)?
Google has flash-lite, too. I liked OpenAi's scheme better, as long as it is parallel and makes sense.
Can anyone give me the compelling reason for Terra and Sonnet to exist?
Virtually every task I give to either is better served by going up a tier and down a reasoning level, or down a tier and up a reasoning level. I simply don't understand their use case. No testing I've done has shown them to have a niche where there isn't a cheaper-with-equal-results or more-performant-at-same-cost option.
I think for the majority of the audience here who use these models inside of coding agent harnesses w/ subsidized monthly subscriptions, there likely isn't one. Depending on how Terra contributes to the rate limits, it may make sense to use it for some subset of a repeated workflow to get more mileage out of sub.
If you pay API prices for inference in your product stack & have specific use-cases w/ custom evals, these smaller models provide an additional cost/speed lever to adjust. Ultimately, I think the point is to compete on cost for use-cases where Sol on the lowest reasoning is overkill.
If you have a fairly straightforward usecase where Terra performs just as well as Sol, but is faster and cheaper, why not use it?
Yes but you are missing the point. It's the case where Luna is not good enough, and Sol is too much. The problem is that on benchmarks Luna xhigh is typically comparable to Sol low in terms of capability. So where does Terra fit in?
1 reply →
Idk… I’ve written 100k sloc in ~2 weeks on the Claude £20 subscription with sonnet medium alone. Seems pretty good to me.
I guess the question is whether you could have done the same using Haiku High/XHigh or Opus Low/Medium at roughly the same speed and cost.
On the API we use a lot of Luna (cost efficient) and sometimes Terra level models. I think we never use a larger model. These are models that our app runtimes depend on.
On subscription for coding and stuff I am basically on Astra Ultra all day long.
They are the biggest models supported in Github Copilot Pro
https://docs.github.com/en/copilot/reference/ai-models/suppo...
You should look at model reasoning level curves on the cost/performance chart on artificial analysis. If you compare the curves for haiku/sonnet/opus I would agree with you that sonnet serves no purpose because it is priced too high. If you look at luna/terra/sol you will see that terra is priced such that it successfully fills a gap between luna and sol. And when you add speed into the equation it adds another dimension where terra makes sense. terra low falls between luna xhigh and max on both cost and intelligence, but it is way faster because it uses way fewer reasoning tokens. So I would say there are workloads where terra makes sense. But from my analysis I don't see any workloads where sonnet makes sense.
There is a chasm of performance difference between haiku and sonnet ime. Sonnet is cheaper and faster than Opus, so if you're providing a service via api where cost and speed is important, it's a big difference.
Except Sonnet is slower, more expensive, and worse at reasoning levels above ~medium. Opus on low is faster, cheaper, and better at doing things.
2 replies →
Sonnet is way faster than Opus when doing everyday tasks, I think on the subscription plans it's a better option when you need something done faster than Opus' long thinking sessions.
For coding, no.
I have some business automation process for some customers that I maintain, where Terra with low reasoning ended up with the best price/performance & latency. With an equivalent Luna with high reasoning, price was usually better, but quality had a higher variance (very much depends on where the thinking takes the model) and latency for high reasoning was worse.
If you trust these benchmarks, I'd argue that there's almost never a reason to use anything other than Terra 6.1 xhigh.
If you use API in agent or your own integration with software, such models make a lot of sense. We use Terra 5.6 for "question to SQL" translation for example. It's good enough and affordable enough. Luna will probably not handle complex queries well while Sol is too expensive.
i use sonnet to generate answers to help-desk tickets, which it does for significantly cheaper and insignificantly worse than opus.
if you're just talking about use in codex or claude code, then sure, leave it on opus all the time. but for running routine api requests, sonnet is great.
I've been refused to do cybersecurity (read-only investigative access against the recent brotli cve) with Opus 5.5 when I had full SSH access, so I guess you've got to have a fallback?
Sometimes this tier is faster, and getting a good enough result in a few seconds instead of a better result in ~15-30 seconds is a meaningful result.
I think it's also a fake lever to give to finops / it teams to try and keep costs down.
I cannot figure it out either. I use Astra Ultra to plan and 6.1Sol High or X-High to build and even on huge projects with 10s of 1000s of pages it is fast and competent
> Can anyone give me the compelling reason for Terra and Sonnet to exist?
To justify the pricing of Sol and Astra. Old business trick.
This is the only reason that makes sense to me. Thank you
I'm on Claude's default plan for personal use, and Sonnet is plenty enough for small projects while not hitting my usage limits too hard.
Cheap and affordable. Dispatching proper tasks to these models can save you a huge % of your quota. Especially if you _know_ the task is simple (ie add something here or there, or archive a bunch of files and reformat kinda thing)
Maybe it's my tasks, but in the cases you're describing, I'm usually better off going down to Luna. I specifically don't see the gap where "It's too complex for Luna-high" and "It's not complex enough for Sol-low" exists.
1 reply →
cool update. Like super cool.
Just one thing about the advert, how can I be the only person here thinking the ad is weird - like for the marketing dep, its a bit weird the ad features the same tram that broke and tragically like ended some people - happened for like the first time ever in 100 years, like a year ago ( 14 months ago) just a bit weird.
Im honestly not trying to be negative, but its kinda weird I kinda feel I should mention it because noone else has?
Im sure literally anyone from southern europe must have thought this the second they saw the ad. It was big news last year.
Am I mad? Just a heads up for PR to always check things like this first when releasing a video? Maybe its just me? Just trying to help
It's a subliminal warning. "On the trolley problem, this model will try to kill you and make it look like an 'accident'."
its just slop bro, welcome to the future
New version of "it's not that deep" dropped.
the only things we've got to actually make a judgement on are vibes and token anxiety. vibes generally indicate that more expensive == better, but more expensive == more token anxiety. so you end up using the most expensive model x effort that keeps your anxiety in check. yeah, maybe terra is great at technical documents. i'm not going to use it because i'd need to spend a couple of days writing technical documents to work out if it was better than sol and then what if it isn't?
is opus 5.5 better than sol 6.1? hmm, i really can't tell. maybe claude is a bit more diligent and identifies more issues up front. but sol has this can do attitude and the codex harness just gets on and does it's job done.
Still can't really forgive any of them for essentially stealing most of this knowledge regardless...
And then talking about other people distilling from them...
I feel like OpenAI hasn’t been as (publicly) hypocritical and petty about their distillation spats as Anthropic has been?
1 reply →
Anyone who has data that is so important to their livelihood should probably make backups.
The results for factual error rate on "difficult" prompts is staggering to me. Particularly the difference between 5.6-Terra and 6.1-Terra. These are wildly different Pareto fronts. Our custom enterprise agent is currently running on 5.6-Terra and it's already very good. I look forward to updating the configuration.
I'll wait for the runebench results!
On AA intelligence index, this fills in the cost per task frontier nicely between 6 Luna and 6.1 sol. So when 6.1 Luna inevitably releases...
The OAI model tango seems to not have an end in sight, for better or worse. Give me more API toys!
Cost v Performance pareto on Runebench, scoring the same as Opus 5.5 for 1/3 of the cost.
These "Pareto frontier" performance vs price graphs have become very popular and I find them mostly uninformative. Pricing is subjective. You chose the position on the X axis, and which measures to use on the Y Axis.
This tier is what I'd actually point opencode at for PR reviews. The top models are too slow and expensive to run on every diff.
is it eligible for data sharing incentive free token quota?
Am I reading the Agents’ Last Exam v1 graph correctly where it's saying GPT Luna Medium performs better than High, but XHigh is better than Medium?
The error bars on eval graphs are commonly like ±5%. It's fairly common to get out-of-order points when the gap between them is only a few percentage points. If we ran it again, or ran it with minor alterations, I'd expect the ordering to be expected. In general, we don't massage any of our eval numbers and for good statistical hygiene we'll report the first run we do, even if later runs show slightly smoother curves.
(I work at OpenAI.)
That is correct. The gpt-6 models are strange.
I am wondering if there is a site with a perf/efficiency chart of the models over time? I am wondering is the pace still speeding up?
Sell it to me. Why would I ever give money for this?
You are paying API prices, and 6.1 Terra gets you 90% of what 6.1 Sol gets you at half the price.
My understanding is API prices are more expensive then monthly plan, is this true?
My plan for OpenAI recently expired and now I’m thinking to test Anthropic ecosystem. My primary use case is programming only.
Should I go with API use or subscription? For coding harness, I jump between codex, OpenRouter, pi, whatever coding harness is trending on that day
3 replies →
These benchmarks made me want to try sonnet 5.5
They say it has "A significant step up on factuality and coding". My browser is underlining "factuality" in red. AI hallucinations or hallucinatiality are still a big problem in my experience.
First known use of factuality was in 1859. It's a "real" word. You can look it up.
https://www.merriam-webster.com/dictionary/factuality ?
Unfortunately it suffers from the same problem as Luna: it will stop at an earliest excuse and no amount of prompting not to stop until X is done will convince it otherwise.
not gonna lie they did really bad and this drained my usage SO FAST. that model used lots of unnecessary skills and read through way more memory files needed.
It literally burned through 2 subs.
If you are confused, that is how this company was always been, remember GPT 4, then they did some weird versioning.
I personally like Terra, it kind of does what I would ask Sol or Astra, but dont have to count tokens as much. And Luna has been amazing tbh, it takes designs from prototype and in record time is implementing it, so that was a huge win.
But yeah, they always had confusing way of naming products and if you are not following news, you will be confused about what it is.
i'm not sure anyone cares about Terra though.
Previous Terra model was so good it was discontinued.
OpenAi needs a model to compete with Opus 5.5, and that's not Terra 6.1.
It could have been Astra 6.1 but they don't have the courage to make it happen.
https://artificialanalysis.ai/#intelligence-efficiency-tabs
According to this, Sol 6.1 x-high is both cheaper and smarter than Terra 6.1 max.
Is it stronger and cheaper than DeepSeek V4.1 Flash 1017?
If not, yawn...
AA has limited 1017 benchmarks so some terras versus 1017 Max: max - 3 points smarter, 50% more expensive xhigh - same score, 25% cheaper high - 1 point dumber, 50% cheaper
Do people use deepseek for real world development or just for building experimental large projects?
Yes, V4.1 Flash 1017 is my daily driver cloud model. Copes fine with what I'd consider difficult brownfield development, like fixing bugs in vLLM prompt cache and coming up with small/sane overlay patches for them. Maybe it's disappointing for people who want to one-shot a game or website from a single prompt; I haven't tried that.
It's nice that their peak-rate hours end quite early in the day UK time.
I use it at work. It's dirt cheap, blazing fast, and is more than enough for most coding work.
If you live in a country with a weak currency, DeepSeek models are definitely a godsend.
You can do real world development with weaker models than Deepseek Flash.
I still use SOTA for the main model though, but also I'm perfectly aware it's mostly out of FOMO.
Deepseek is more than enough for default subagents though. Any well defined and verifiable task and there's absolutely zero difference between SOTA and good-enough models. Used to be just speed, but Deepseek is fast.
1 reply →
It's only a little better than Deepseek with Max effort in benchmarks. Doesn't bode too well for 6.1 Luna.
The benchmarks are already saturated and the curves for Terra stick to the axes.
Underwhelming
Really hate how they only show it vs Opus @ Max and not Low/Medium/High.
That terminal bench 4.0 result is... bad? Terra XHigh more expensive than Sol Low and still does worse.
It's very likely this is a next checkpoint of GPT-6 "Sol", over-optimised for the web chat and offered to Codex subscribers as a "please go away" offering.
If this was half the price, I'd be very interested, but at 1/5/0.05? Feels like another panic release because they can't get Astra 6.1 out of the door...
Scoring "bad" on Terminal Bench 4.0 can also be a sign that it's cheating less: https://preferencemodel.com/blog/terminal-bench-4-is-saturat...
[dead]
[dead]
[dead]
[dead]
[dead]
People keep asking OpenAI for a mode that's as good as Opus 5.5, and they release this instead. Interesting strategy. Let's see how it works out for them.
They don't have anything competitive with Opus 5.5 yet. There is a supposed model in the works, "Bel" last time I paid attention to gossip, that is thought to be better, later this year. Allegedly Anthropic caught OpenAI off guard and / or their different training approaches ended up working better than anticipated.
My expectation is Fable 5.5 is next overall improvement, and my bet is OpenAI's "Bel" will be the next competitive one (or, their lack of safety culture will bite them/us in the process).
There’s also a chance they surrender the dev market altogether to Ant and just focus on the “normie” segment with dots, computer use, excel etc. In which case they won’t have anything competitive to opus-5.5 anytime soon for SWE.
Luna and Sol-6.1 are crazy cost effective and all flows here have switched to them, so I'd say really well so far.
There's very few stuff that needs Astra/Opus level intelligence.
Luna 6.1 would be amazing.
This release is interesting for enterprise at API rates.
At least unless the Terminal Bench 4.0 score is more indicative than the other ones, because that one shows it clearly worse than 6.1 Sol for both performance and cost.
Taking the benchmarks in the article at face value, this is ~on par with Opus 5.5 for ~1/4 the price!
Which is why I don't take the benchmarks at face value...
4 replies →
Taking the benchmarks (as reported by OpenAI) at face value, why wouldn't they show Opus 5.5 on the OS world benchmark, and show no Anthropic models at all on the "factual error rate" benchmark
...no it's not? At max effort, it scores 50% on terminal bench vs. Opus 5.5's 70%.
2 replies →
I hate Opus 5.5, constant overreach and over-engineering for any project
That's wild to say when comparing to the OpenAI family of models. I've never seen so many tests that validate tests that validate tests.
1 reply →
Medium or High? Depending on the task, you might want to switch between them.
You’re holding it wrong.
"better writing"
let's see if it can pass the Pangram test. Many services purport to fool it, but fail.
Because Pangram is snake oil.
2025 called and it wants it's take on AI detection back.
https://www.pangram.com/blog/third-party-pangram-evals
Note that while that page is from Pangram, the studies on it are not. If they are excluding studies that show Pangram performs poorly in realistic conditions, please share those studies!
If Pangram was snake oil (it's not but could be more mindfully marketed), it's definitely not now that every major lab has added watermarking.
You wouldn't even need to be able to read the information the watermark encodes to recognize it exists in a piece of text
2 replies →
Fooling Pangram and writing well are not the same axes, even if there’s some correlation.
Everyone defaults to checking pangram when AI is susepcted, so it's still an important test.
If they had any courage, they would release 6.1 Astra, but they don't.
Update: Look at the OpenAI employees downvoting me.
The next model will likely be Nova
GPT-6 only just came out! The AI landscape is always on the move!
2026-09-03 is almost 2 months ago already - that's basically half a lifetime in LLM land!
True!