POWERFUL NOUS FINE-TUNE OF THE 34B YI MODEL (HUGGINGFACE HUB)

Summary from the TLDR AI newsletter

Nous research has released a very strong 34B model built on top of the bilingual Yi model. Capybara is trained primarily on multi-turn data with more than 1000 tokens per conversation, which makes it uniquely positioned for certain user-facing tasks with high information density.