LLM Fashion Arena

All evals are fake, but some are useful

---

See:

• All models are wrong but some are useful by James Clear

• The map is not the territory by Farnam Street

• All evals are fake but some are useful by Anthropic’s Logan Graham

---

What is the fashion style for ChatGPT, o1, Claude, Qwen, or Figure? What does fashion signal about the underlying model? :)