QVQ-72B-Preview
Experimental 72B model that reasons over images
About
QVQ-72B-Preview is the Qwen team's experimental take on visual reasoning: Qwen2-VL-72B extended with o1-style step-by-step thinking over images and text. The team reports 70.3 on MMMU and 71.4 on MathVista, edging past o1 on the latter. It is explicitly a research preview — the Qwen team lists language mixing, recursive reasoning loops, and lost focus on long visual chains as known issues — so it suits experimentation more than production. Weights are open under the custom Qwen license, which is not OSI-approved; commercial users should check the terms.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Qwen-reported: MMMU 70.3, MathVista (mini) 71.4, MathVision 35.9
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: Qwen team release Dec 24 2024 confirmed via Qwen announcement and HF README: 72B visual reasoning on Qwen2-VL-72B base.; Benchmarks Qwen-reported: MMMU 70.3, MathVista 71.4, MathVision 35.9.; Known issues (language mixing, reasoning loops) acknowledged by Qwen team — documented as limitations.; License caveat: custom Qwen license, not OSI-approved — flagged..