signalAI热榜2026-10-04
LMSYS Releases Open-Source Chatbot Model Vicuna-13B, Fine-Tuning LLaMA with ShareGPT Conversations, Training Cost About $300
Vicuna-13B is an open-source chatbot fine-tuned on about 70K user-shared conversations from ShareGPT, with a training cost of about $300. A preliminary GPT-4 evaluation shows it achieves over 90% of ChatGPT/Bard quality, outperforms LLaMA and Alpaca in over 90% of questions, and is equal or better than ChatGPT in 45% of cases. The code, weights, and online demo are released for non-commercial use.
- for who
- AI researchers and developers interested in open-source chatbots and low-cost fine-tuning.
- why now
- Vicuna-13B's March 30 release offers near-ChatGPT quality, open weights, and $300 training cost.
- what changes
- They can access a high-quality open-source chatbot and replicate its training and evaluation methods at a fraction of the cost.
- to do
- Use the released code and weights for non-commercial projects or explore the GPT-4-based evaluation framework.
key points
- Vicuna-13B fine-tuned on 70K ShareGPT conversations, cost about $300
- GPT-4 evaluation: over 90% quality of ChatGPT/Bard, better than LLaMA and Alpaca in most tests
- Open-source release includes code, weights, and online demo under non-commercial license
#Vicuna-13B#open-source model#fine-tuning#LLaMA#low-cost training
score
score 9 out of 10. 0-10: how dense the facts are, multiplied by how much you can do with them after reading. 8+ means the topic's evidence bar is met: benchmarks and availability for a new model, amount and investors for a funding round, revenue figures for a solo-money story. Below 5 an item does not enter the digest. A press release scores 3 or less, a reprint loses 2, anything older than 14 days loses 1, a headline that misleads loses 3.
read the source