Spaces:

---

open-llm-leaderboard

/

open_llm_leaderboard

like 12.7k

Running on CPU Upgrade

There clearly something wrong with test when mistral 7b and Qwen 2.5 72B have eval error. Is this what you are looking in to or for people just forget about trying to use leaderboard for models that failing? (even if they loads correctly locally) Uncertainty is unnerving. Thx for your work.