LLaVA Project.

Open-source multimodal model linking vision encoders to LLMs. UW Madison + Microsoft collab.

Models from LLaVA Project.

LLaVA 34B

34B

by LLaVA Project · LLaVA · 4,096 ctx

Largest open-weight LLaVA — vision encoder + Yi-34B backbone.

LLaVA 13B

13B

by LLaVA Project · LLaVA · 4,096 ctx

LLaVA 7B

7B

by LLaVA Project · LLaVA · 4,096 ctx