Vicuna

LMSYS's chatbot fine-tuned on ShareGPT conversations

Code & DevelopmentFree

Vicuna is an open-source chatbot model from LMSYS Org fine-tuned from LLaMA on ShareGPT user conversations. When released, Vicuna-13B achieved 90%+ of ChatGPT's quality on human evaluations at a fraction of the cost. Vicuna introduced the practice of fine-tuning on real human-ChatGPT conversations, establishing a template that dozens of subsequent models followed. It remains widely used as a base for research and evaluation benchmarking.

Key Features

• ✓ShareGPT fine-tuning
• ✓90% ChatGPT quality
• ✓7B/13B/33B sizes
• ✓Research benchmark
• ✓Open weights
• ✓llama.cpp compatible