UC Berkeley's open-source Vicuna LLM chatbot released new improved model weights

> Users have to apply it on top of the original LLaMA weights to get actual Vicuna weights.

Given the Apache 2 release of Cerebras-GPT-13B (https://huggingface.co/cerebras/Cerebras-GPT-13B) is LLaMA still that much better to bother with all that non-free "yes, but patch it first" scheme?

---

ed: it also seems they've integrated Dolly-v2-12b in their demo, if one wants to have a play with it https://chat.lmsys.org/

Twitter announcement: https://twitter.com/lmsysorg/status/1646618380114468864

Interested to see how it compares to copilot or chatgpt for code competition