3 ms·

Not surprising, but damn. $60B to $170B in ~5 months?

by paxys 1y ago

aurareturn 1y ago
https://openrouter.ai/rankings https://openrouter.ai/rankings They've been dominant. ChatGPT is still the better consumer AI tool but Anthropic models have been better for professionals. When I use Github Copilot agent mode, Sonnet 4 is worlds better than GPT 4.1 and Gemini 2.5 Pro.
- sigmoid10 1y ago
  They are absolutely not dominant. OpenRouter only captures a tiny fraction of the market. I think around 2-3 million users. But Claude has something like 10-30 million active users according to most estimates. And ChatGPT is already north of 300-400 million active users. Valuations in this sector have nothing to do with facts anymore, because otherwise noone would give Anthropic a valuation in the same order of magnitude as OpenAI for a tiny fraction of the user base while all objective benchmarks put the top models so close together. This is pure insanity riding on a vague hope that one of those companies will capture AGI while the others drop out.
  - aurareturn 1y ago
    OpenRouter only captures a tiny fraction of the market. I think around 2-3 million users. But Claude has something like 10-30 million active users according to most estimates. And ChatGPT is already north of 300-400 million active users In case it wasn't obvious, I implied the same thing. Anthropic models are more popular on OpenRouter, which I take it as it's more preferred for developers and their apps. However, ChatGPT is still better.
    - sigmoid10 1y ago
      This implies that all (or at least most) developers use openrouter, which is a very bold and unlikely statement given the above numbers.
      - aurareturn 1y ago
        Nope. I did not imply that. Not sure how you're getting so many assumptions wrong. I'm using OpenRouter as a proxy for how developers are seeing models.
        sigmoid10 1y ago
        And I'm saying that's not a valid proxy in any meaningful way - for developers or anyone else. Not sure how you're not getting that given the available numbers.

Not surprising, but damn. $60B to $170B in ~5 months?

by paxys 1y ago

Not surprising, but damn. $60B to $170B in ~5 months?

https://openrouter.ai/rankings https://openrouter.ai/rankings They've been dominant. ChatGPT is still the better consumer AI tool but Anthropic models have been better for professionals. When I use Github Copilot agent mode, Sonnet 4 is worlds better than GPT 4.1 and Gemini 2.5 Pro.

sigmoid10 1y ago
They are absolutely not dominant. OpenRouter only captures a tiny fraction of the market. I think around 2-3 million users. But Claude has something like 10-30 million active users according to most estimates. And ChatGPT is already north of 300-400 million active users. Valuations in this sector have nothing to do with facts anymore, because otherwise noone would give Anthropic a valuation in the same order of magnitude as OpenAI for a tiny fraction of the user base while all objective benchmarks put the top models so close together. This is pure insanity riding on a vague hope that one of those companies will capture AGI while the others drop out.
- aurareturn 1y ago
  OpenRouter only captures a tiny fraction of the market. I think around 2-3 million users. But Claude has something like 10-30 million active users according to most estimates. And ChatGPT is already north of 300-400 million active users In case it wasn't obvious, I implied the same thing. Anthropic models are more popular on OpenRouter, which I take it as it's more preferred for developers and their apps. However, ChatGPT is still better.
  - sigmoid10 1y ago
    This implies that all (or at least most) developers use openrouter, which is a very bold and unlikely statement given the above numbers.
    - aurareturn 1y ago
      Nope. I did not imply that. Not sure how you're getting so many assumptions wrong. I'm using OpenRouter as a proxy for how developers are seeing models.
      - sigmoid10 1y ago
        And I'm saying that's not a valid proxy in any meaningful way - for developers or anyone else. Not sure how you're not getting that given the available numbers.