Atlas

AI labs

AI labsUnited States

LLaVA Authors models

The LLaVA research collaboration develops open-weight vision-language models and evaluation resources. It publishes model artifacts, datasets, papers, and leaderboards for the LLaVA and LLaVA-OneVision families.

llava-vl.github.io

Scored models
3
Best ranked
LLaVA-OneVision 72B #531 of 896
Newest
LLaVA-OneVision 72B Aug 6, 2024
Open weights
3 of 3

LLaVA Authors models ranked by capability

Capability is Atlas's single score for each model, fitted to every sourced benchmark result (methodology). Rank is among all 896 ranked models from every lab. The 1 model without enough scores for a capability estimate follow, newest first.

ModelReleasedCapabilityRankBenchmarksWeights
LLaVA-OneVision 72BAug 6, 2024118.7#53113Open
LLaVA v1.5 7BOct 5, 202381.7#8387Open
LLaVA v1.6 Mistral 7BJan 30, 2024——2Open

All AI labsOpen-weight LLM leaderboard