It shouldn't be surprising OpenAI does have the most compute out of all the major labs. The only reason why Anthropic models are expensive is they are the most in demand models in the world and Anthropic is fighting for compute. The only way to you limit demand for your model is increasing API pricing this is also why Anthropic probably has great margin and probably is profitable compared to OpenAI.
I doubt there's any sort of criminal behavior there - the model is anthropic's up and anthropic probably charges a very expensive license fee that's the same for all of them, and their cogs on compute aren't going to be wildly different, so the main drivers of the cost are roughly the same and they're all offering customers the same end product so the prices would likely also be similar in the end
There is a blog post waiting to be written (that I won't write) about the size/effort tradeoffs, and particularly how small models get some surprisingly good results with lots of turns and reasoning.
DeepSWE will let you chart turns taken or tokens used, and FrontierCode will chart tokens. If you use that, you can see Sol high and Terra max get about the same DeepSWE number, but Terra max takes twice the turns. Luna max scores a smidgen lower with even more turns.
Smaller models relying on lots reasoning may "scale down" better on easier tasks, because unlike size, reasoning effort is dynamic: the model can see the task looks easy and stop. On DeepSWE, the cost curves for the three 5.6 models are almost on top of each other, but on FrontierCode Extended, the version of FrontierCode with the most everyday tasks in the mix, there's a spread of costs at the ~55% level.
The recent Laguna S 2.1 model (118B, 8B active) puts up surprising coding numbers for its size, and the lab behind it specifically credits its "way of working (persistence, verification, willingness to backtrack)". Some other open models that folks report getting good mileage out of seem to get there partly by throwing a lot of reasoning at the problem.
There is a little bit of a question, if some models rely on getting it wrong a bit more at first and external checks catching the problems, of whether they're also more frequently getting things wrong they can't self-verify (say, quality of UI or API design) and then it falls to the human to find it. Still, getting the results they're getting at all is neat.
Some benchmarks historically favored reporting only on the max variants, maybe because they want to show the frontier? but that is not always what you need for practical decision. (AA has the full effort sweep for Opus 5 and Sol/Luna/Terra at least.) And at least FrontierCode finds Opus 5 taking a hit in performance above 'medium'.
I am not trying to pick a winner here. I'm probably not going to use tiny models on max for everything, but I think it's cool that you can get so much more out of a small model by amping up reasoning, tool use, and persistence.
Forgot about the ol "but how many tokens did you spend to get _there_"--wish benchmarks would include the number of input/output tokens to achieve the score. I think the closest is Android Bench https://developer.android.com/bench although best you can do is extrapolate off time/cost (iirc they claim to prefer using provider's native API)
In general, smart models work fine with any tools, dumber models need better tools to achieve same results but better tools can eat more context/take more turns
I've gotten decent results with Llama 3.1 8b on Hugging Face tester with Exa MCP since it seems to dump sufficient context into WebSearch/WebFetch type calls even a crappy model almost always gets back what it needs as long as it calls the tool at least once. I had Claude Code look at previous sessions with SearXNG vibe MCP compared to Exa MCP and results got better when it modified SearXNG to work very similarly to Exa. Ended up with this https://gist.github.com/nijave/604c43e3e0fdcd60f5280d3a6b109... although it's really only optimized for "search" not "fetch" at this point. Fetch is basic niquests without Javascript or anything clever
It shouldn't be surprising OpenAI does have the most compute out of all the major labs. The only reason why Anthropic models are expensive is they are the most in demand models in the world and Anthropic is fighting for compute. The only way to you limit demand for your model is increasing API pricing this is also why Anthropic probably has great margin and probably is profitable compared to OpenAI.
Not just compute for OAI, GPT-5.6 is more token efficient across the board vs the Anthropic equivalents: https://artificialanalysis.ai/models?intelligence-index-toke...
No wonder why Tibo can afford to hit the reset button liberally.
I also suspect there is a price fixing agreement between all of the inference providers for Claude (such as Amazon, Anthropic, Microsoft, etc).
I doubt there's any sort of criminal behavior there - the model is anthropic's up and anthropic probably charges a very expensive license fee that's the same for all of them, and their cogs on compute aren't going to be wildly different, so the main drivers of the cost are roughly the same and they're all offering customers the same end product so the prices would likely also be similar in the end
8 replies →
"Price fixing" isn't the correct term here but yes, it's very common to have the same price across different retailers/resellers.
1 reply →
I think on swebench verified luna was only like 3% points lower for 1/5 the cost
Like 96% vs 93% or something
There is a blog post waiting to be written (that I won't write) about the size/effort tradeoffs, and particularly how small models get some surprisingly good results with lots of turns and reasoning.
DeepSWE will let you chart turns taken or tokens used, and FrontierCode will chart tokens. If you use that, you can see Sol high and Terra max get about the same DeepSWE number, but Terra max takes twice the turns. Luna max scores a smidgen lower with even more turns.
Smaller models relying on lots reasoning may "scale down" better on easier tasks, because unlike size, reasoning effort is dynamic: the model can see the task looks easy and stop. On DeepSWE, the cost curves for the three 5.6 models are almost on top of each other, but on FrontierCode Extended, the version of FrontierCode with the most everyday tasks in the mix, there's a spread of costs at the ~55% level.
The recent Laguna S 2.1 model (118B, 8B active) puts up surprising coding numbers for its size, and the lab behind it specifically credits its "way of working (persistence, verification, willingness to backtrack)". Some other open models that folks report getting good mileage out of seem to get there partly by throwing a lot of reasoning at the problem.
There is a little bit of a question, if some models rely on getting it wrong a bit more at first and external checks catching the problems, of whether they're also more frequently getting things wrong they can't self-verify (say, quality of UI or API design) and then it falls to the human to find it. Still, getting the results they're getting at all is neat.
Some benchmarks historically favored reporting only on the max variants, maybe because they want to show the frontier? but that is not always what you need for practical decision. (AA has the full effort sweep for Opus 5 and Sol/Luna/Terra at least.) And at least FrontierCode finds Opus 5 taking a hit in performance above 'medium'.
I am not trying to pick a winner here. I'm probably not going to use tiny models on max for everything, but I think it's cool that you can get so much more out of a small model by amping up reasoning, tool use, and persistence.
Forgot about the ol "but how many tokens did you spend to get _there_"--wish benchmarks would include the number of input/output tokens to achieve the score. I think the closest is Android Bench https://developer.android.com/bench although best you can do is extrapolate off time/cost (iirc they claim to prefer using provider's native API)
In general, smart models work fine with any tools, dumber models need better tools to achieve same results but better tools can eat more context/take more turns
I've gotten decent results with Llama 3.1 8b on Hugging Face tester with Exa MCP since it seems to dump sufficient context into WebSearch/WebFetch type calls even a crappy model almost always gets back what it needs as long as it calls the tool at least once. I had Claude Code look at previous sessions with SearXNG vibe MCP compared to Exa MCP and results got better when it modified SearXNG to work very similarly to Exa. Ended up with this https://gist.github.com/nijave/604c43e3e0fdcd60f5280d3a6b109... although it's really only optimized for "search" not "fetch" at this point. Fetch is basic niquests without Javascript or anything clever
Luna is the most impressive model released so far by any provider. It's perfect for doing all the low-level tool calling and developing hypotheses.
Terra is great for the humans to talk to.
Sol is really only useful if you need to do more delicate things like synthesis of multiple competing pieces of information.
A system that uses all three variants will massively outperform a system that just uses the biggest model for everything.
Yeah, sol is impressive but IMO Luna is the real standout (and terra is the laggard of the group) for performance/cost
This must be on API costs, not counting the $100/200 tiers, right?
yes; fyi usage limits on the $200 claude sub correspond to at least $1.2k/week in api tokens
More than that using Fable!
Probably because they made ASICs to run inference for less.
Are those actually deployed at scale yet?
Yes.
1 reply →