LLaVA-NeXT
Open-source multimodal model that beat Gemini Pro on image benchmarks.
About
LLaVA-NeXT (also called LLaVA-1.6) was released on January 30, 2024, by the LLaVA academic team, as an open-source large multimodal model trained on text-image data. Its AnyRes technique, which represents high-resolution images as multiple image crops, boosted reasoning, OCR, and world knowledge, and the 34B variant exceeded Gemini Pro on several image benchmarks including MMMU and MathVista.
The model demonstrated surprisingly strong zero-shot video understanding, leading to the LLaVA-NeXT-Video variant. Code, checkpoints, and training recipes were released openly, making it one of the most reproduced open vision-language models of 2024.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 75; Trending score 74/100 from researcher assessment.