Qwen3-VL-235B-A22B-Instruct: Alibaba's Largest Qwen3 Vision-Language Model

What Qwen3-VL-235B-A22B-Instruct is

Qwen3-VL is Alibaba's dedicated vision-language line, and this is its largest instruct release: 235 billion total parameters with 22 billion active per token, Apache 2.0 licensed and self-hostable, with a 256K token context window. Alibaba publish the family in both dense and mixture-of-experts architectures scaling from edge devices to cloud deployment, and in two editions per size — Instruct, which is this one, and a reasoning-enhanced Thinking edition covered separately here.

Qwen3.5 and the Qwen3-VL line

Alibaba's Qwen3.5 generation is built on a unified vision-language foundation using early fusion training on multimodal tokens, and Alibaba state that it reaches parity with Qwen3 while outperforming the Qwen3-VL models across reasoning, coding, agents and visual understanding.

LMArena

On LMArena's text leaderboard Qwen3 VL 235B A22b Instruct has 1413 Elo (rank 148); by category, 1463 Elo in Coding (rank 137), 1409 Elo in Math (rank 153), 1442 Elo in Expert (rank 133) and 1356 Elo in Creative Writing (rank 175). The Thinking edition has 1395 Elo (rank 171, thinking effort). LMArena's text board does not show the model an image.

Price

Roughly $0.21 per million input tokens through third-party providers, an aggregated figure rather than first-party.

What Alibaba list for Qwen3-VL

Alibaba list the Qwen3-VL generation's capabilities as: a visual agent that operates PC and mobile interfaces, recognising elements, understanding functions, invoking tools and completing tasks; visual coding that generates Draw.io diagrams and HTML, CSS and JavaScript from images and video; spatial perception covering object positions, viewpoints and occlusion, with 2D grounding and 3D grounding; long-context and video understanding, with a native 256K context expandable to 1M and second-level indexing over hours-long video; broader visual recognition of celebrities, anime, products, landmarks, plants and animals; and OCR in 32 languages, up from 19, with better handling of low light, blur, tilt, rare characters and long-document structure. Alibaba describe its text understanding as on par with pure language models.

Architecture changes

Alibaba describe three architectural changes. Interleaved-MRoPE allocates positional frequencies across time, width and height for long-horizon video reasoning. DeepStack fuses multi-level ViT features to capture fine detail and sharpen image-text alignment. Text-timestamp alignment replaces T-RoPE with timestamp-grounded event localisation for video. The model loads through the Qwen3VLMoeForConditionalGeneration class in Transformers, which Alibaba advised building from source before version 4.57.0 was released, and Alibaba recommend flash_attention_2 for multi-image and video inputs.

Release and settings

Alibaba released Qwen3-VL-235B-A22B in Instruct and Thinking editions on 23 September 2025. The 30B-A3B models and FP8 versions across the collection followed on 4 October, the 4B and 8B models on 15 October, and the 2B and 32B models on 21 October, each in Instruct and Thinking editions; the technical paper was published on 27 November 2025. Qwen3-VL is supported in vLLM 0.11.0 and later, and Alibaba's efficient H100 and H200 deployment uses tensor parallelism of 8 with expert parallelism enabled. For Instruct models Alibaba recommend temperature 0.7, top_p 0.8, top_k 20, a presence penalty of 1.5 and up to 32,768 output tokens.

Lineage

Qwen3-VL follows Qwen2.5-VL, Qwen2-VL and the original Qwen-VL, each with a technical report on arXiv.

Official Resources