qwen
Qwen: Qwen3 VL 235B A22B Instruct
qwen/qwen3-vl-235b-a22b-instruct
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
Qwen3-VL-235B-A22B-Instruct is an open-weight mixture-of-experts vision-language model from Qwen that accepts text and image inputs and returns text. Its model card describes it as a general vision-language system covering VQA, document parsing and chart understanding, plus vendor-reported strengths in GUI agent control, spatial grounding, video comprehension and OCR across 32 languages. OpenRouter lists a 262,144-token context, tool calling and structured outputs; the card claims 256K native context expandable to 1M. Weights are Apache-2.0 licensed. No independently measured performance data was available in this pack.[1][2]
Strengths
- Vendor-reported multimodal breadth: GUI agent operation on desktop and mobile, 2D/3D spatial grounding, video understanding with timestamp-level event localization and OCR in 32 languages.[2]
- Long context per listing and card: OpenRouter shows a 262,144-token window, and the card claims native 256K expandable to 1M for book-length documents and hours-long video.[1][2]
- Open-weight Apache-2.0 release; OpenRouter lists tool calling, structured outputs and text-plus-image input with text output.[1][2]
Limitations
- Evidence gap: this pack holds no independently measured benchmark results; the card's performance tables are images without readable figures and OpenRouter's benchmark fields are empty. Validate vendor capability claims on your own workloads.[1][2]
- Self-hosting friction: the card advises building transformers from source because the released version did not yet include this architecture, and recommends FlashAttention-2 for memory savings at this 235B-class MoE scale.[2]
- Modality mismatch to verify: OpenRouter lists only text and image inputs while the card describes video understanding, so video on that route may need to reach the model as sampled frames.[1][2]
Where to use it
- Editorial suggestion: document intelligence pipelines using OCR in 32 languages, long-document structure parsing and chart or table question answering, per vendor claims.[2]
- Editorial suggestion: long-context vision assistants for book-length files or extended video review where the 256K-plus context is the draw; validate recall quality first.[1][2]
- Editorial suggestion: prototyping GUI agents or image-to-code generation such as Draw.io, HTML, CSS and JS, both described by the vendor but unmeasured here.[2]
Sources
- [1] OpenRouter catalog · Qwen: Qwen3 VL 235B A22B InstructSources checked: Nov 13, 2026
- [2] Model documentation · Qwen: Qwen3 VL 235B A22B InstructSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
Qwen3-VL-235B-A22B Instruct is an open-weight multimodal model that unifies strong text generation with visual understanding across images and video. The Instruct model targets general vision-language use (VQA, document parsing, chart/table...
Specifications
- Model ID
- qwen/qwen3-vl-235b-a22b-instruct
- Creator
- qwen
- Added to OpenRouter
- 23.09.2025
- Context length
- 262,144
- Modality
- text+image->text
- Input modalities
- text, image
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.210 / out $1.9
- Knowledge cutoff
- 31.03.2025
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | No measurement | — |
| Coding | No measurement | — |
| Agents | No measurement | parameters: structured_outputs, tool_choice, tools |
| Writing | No measurement | output: text |
| Reasoning | No measurement | — |
| Document analysis | No measurement | — |
| Vision | No measurement | input: image |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
No evaluations are supplied for this model.
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| DeepInfrafp8 | 262,144 | $0.200 | $0.880 | — | — | 100% |
| Venicefp8 | 128,000 | $0.210 | $1.9 | — | — | 100% |
| Parasailfp8 | 131,072 | $0.210 | $1.9 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.