Molmo 2
Fully open video-understanding VLM with spatio-temporal grounding
About
Molmo 2 is Ai2's second-generation vision-language family, with weights released in early 2026 in 4B, 8B, and a fully-open 7B-O variant built on OLMo. It extends the original Molmo from images to video: it tracks objects across frames, pinpoints when and where events happen, and does dense captioning and counting by pointing. Ai2 reports the 8B beating Gemini 3 Pro on video pointing and tracking, with 85.7 on TextVQA and 93.2 on DocVQA. Weights, datasets, and training code are all public under Apache 2.0, though Ai2 warns some third-party training datasets carry non-commercial restrictions. Long video and dense-document OCR remain weaker spots.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
TextVQA 85.7, DocVQA 93.2; video pointing/tracking ahead of Gemini 3 Pro (Ai2-reported)
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 (verified tier, metrics-v1). Evidence: Weights released early 2026 (announced Dec 2025); 4B/8B/7B-O variants, ~37K context verified via OpenRouter and Ai2 coverage; TextVQA 85.7, DocVQA 93.2, video grounding ahead of Gemini 3 Pro are Ai2-reported; labeled as such; Apache 2.0 weights but third-party training datasets carry non-commercial constraints; layered-license caveat captured.