I thought the same about xAI / Grok, and was surprised to see the Grok Code Fast 1 at the top of the list for number of reasoning tokens processed in Open Router’s state of AI report that came out this month:
I'm curious about Grok - is there a different version than what is used as the chatbot on X? I see a lot of ridiculous responses that people get from Grok chatbot, so it gives the impression that it isn't very good.
https://openrouter.ai/state-of-ai#reasoning-models-now-repre...