3 ms·

Anthropic is known for having a model with a context nearly 3x that of ChatGPT. You must not be following LLMs very closely

• benreesman 3y ago
  As someone who had been rabidly following developments in the space until earlier this year: I’ve fallen a bit behind. I’be been walking around with the (seemingly misguided) view that the bigger LLaMA 2 models were sort of neck and neck with like 3.5-turbo, Claude, and the rest of the first string modulo GPT-4 that had a pretty steady if eroding lead. What’s the lay of the land in 9/23? Or better yet, what do I need to bookmark to get the inside baseball?
  • p1esk 3y ago
    Don’t worry, GPT-4 is still the best.
    • benreesman 3y ago
      Water cooler chat is that it’s a 1.5 terraweight, 16-way mixture model that is running at a loss on inference for paying customers. If that’s true it fucking better be.
      • p1esk 3y ago
        This should not have stopped any of FAANGs from competing (especially Google), and yet they are still far behind OpenAI - more than a year after GPT-4 was trained.