Comment by avazhi
10 hours ago
I have no personal skin in the game in terms of investment posture, but OpenAI is, by an increasing margin, the weakest player. Claude and Gemini are both blatantly better (better as in smarter/more capable across all measures). Claude seems like the ‘smartest’ model and while Gemini is way more annoying to interact with in terms of its sycophantic nonsense and brain rot writing style, Google also has unlimited compute and I’ve literally never run out of tokens using any of Gemini’s models. And meanwhile Anthropic is seemingly addressing its biggest weakness, which is limited compute, by basically taking over from Grok’s computer hardware (I half expect Grok to get discontinued any day now - it sure seems like xAI has accepted that Claude is the front runner and they’re just getting behind it, kind of like what OpenAI agreed to do if they ever got behind in the AGI race back in ~2017).
So what does OpenAI even lead at? Name recognition because they were first? At some point they were supposed to be specialising in medicine but I notice no difference between Gemini and ChatGPT when it comes to medical questions or analysis.
My prediction is OpenAI will be the first big one to go bankrupt or be acquired, which is also probably why they are rushing this IPO: gotta get the founders cashed out.
Somewhat of an aside, but I have no idea if AGI is actually possible with LLMs, but Claude is the closest thing to a person that I’ve used (even if it has its moments of abject retardation - not unlike humans, I guess).
Honestly, even if anthropic models are better than OpenAIs, I don't understand how they want to make money. In October last year, the frontier models from US companies were so much better than the cheapest models, I thought it was over, but nowadays, even smaller models perform adequately enough. I now use free models (or rather, very cheap ones) in all of my personal projects, and even though anthropic's harness is hard to replace on complex cases (it helps understand where the LLM failed better, which allows to correct the mistakes more easily), I'm pretty sure pure LLM gains are less and less with each new models.
> I have no personal skin in the game, but OpenAI seems like the weakest player. Claude and Gemini are both blatantly better.
The market doesn't necessarily reward better products or (in this case) more intelligence.
If it did, I'd be a lot richer than many of the mainstream startups.
> The market doesn't necessarily reward better products or (in this case) more intelligence.
It does when the product being sold is sold based on how intelligent (and thus how capable) it is. Unfortunately with people intelligence is merely an imprecise proxy of capability or organisational productivity.
Anthropic models are well used for coding and similar tasks, and mostly through their own tooling as they are pretty aggressive limiting other usage.
But I don't see their models being used that much through api for all the applications that are using api nowadays. Openai is the one with the easiest api to use and the more lax about it.