Qwen2-VL
Open vision-language models that read documents, watch video, and operate phone UIs.
About
Alibaba's Qwen team open-sourced Qwen2-VL on August 30, 2024 (2B and 7B; the 72B flagship followed on September 19) with a new M-RoPE positional encoding that handles 1D text, 2D images, and 3D video in one scheme. Team-reported results led DocVQA, RealWorldQA, and MTVQA at launch. It understands images at any resolution, follows videos longer than 20 minutes, reads multilingual text in images, and can drive smartphone agents. Licensing is split: 2B and 7B are Apache 2.0, the 72B uses the restrictive Qwen License. Honest caveats: the phone/robot agent demos are lab showcases, not shipped products; and the very best vision scores belong to the 72B most people won't self-host.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Team-reported: state-of-the-art on DocVQA, RealWorldQA, and MTVQA at release (Aug 2024).
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: Release 2024-08-30 (2B/7B) and 2024-09-19 (72B) verified via Qwen GitHub news section.; Apache 2.0 for 2B/7B, Qwen License for 72B — verified in repo README.; Team-reported SOTA on DocVQA, RealWorldQA, MTVQA at launch — labeled as team-reported.; Agent-demo caveats and 72B license cap disclosed..