qwen
Qwen: Qwen3 235B A22B Instruct 2507
qwen/qwen3-235b-a22b-2507
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
Qwen3-235B-A22B-Instruct-2507, listed as qwen/qwen3-235b-a22b-2507, is a multilingual instruction-tuned mixture-of-experts model with 235B total and 22B active parameters, and a native 256K-token context extendable to about 1M via a specialized serving setup. The vendor reports gains over the earlier Qwen3 non-thinking variant in reasoning, coding, alignment, multilingual ability, and long-context understanding; those benchmark results are largely self-reported. OpenRouter lists it as text-in/text-out with tool calling and structured output support. It runs only in non-thinking mode.[1][2]
Strengths
- In the vendor's own evaluations, this variant scores markedly higher than its non-thinking predecessor across reasoning, mathematics, coding, and multilingual suites, staying competitive with frontier peers.[2]
- Native 256K context window, extendable to roughly 1M tokens through documented sparse-attention serving modes, with strong reported long-context accuracy.[1][2]
- Supports tool calling and structured outputs; the vendor recommends agentic use through its Qwen-Agent framework.[1][2]
Limitations
- Runs only in non-thinking mode and does not emit thinking blocks, which may matter for workflows that expect explicit reasoning traces.[2]
- Text-only input and output; no vision or audio modalities are listed.[1]
- Most capability and benchmark evidence is vendor-supplied; going beyond 256K tokens needs custom configuration and about 1000 GB of GPU memory per the model card, so workload-level validation is advised.[1][2]
Where to use it
- Analysis and summarization of very long documents, where the extended context may help; validate retrieval quality on your own corpus first.[2]
- Tool-using agents and function-calling workflows, for example via the vendor's Qwen-Agent framework.[1][2]
- Multilingual writing, reasoning, and coding assistance for teams seeking a self-hostable instruct model.[2]
Sources
- [1] OpenRouter catalog · Qwen: Qwen3 235B A22B Instruct 2507Sources checked: Nov 13, 2026
- [2] Model documentation · Qwen: Qwen3 235B A22B Instruct 2507Sources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
Qwen3-235B-A22B-Instruct-2507 is a multilingual, instruction-tuned mixture-of-experts language model based on the Qwen3-235B architecture, with 22B active parameters per forward pass. It is optimized for general-purpose text generation, including instruction following,...
Specifications
- Model ID
- qwen/qwen3-235b-a22b-2507
- Canonical ID
- qwen/qwen3-235b-a22b-07-25
- Creator
- qwen
- Added to OpenRouter
- 21.07.2025
- Context length
- 262,144
- Modality
- text->text
- Input modalities
- text
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.140 / out $0.800
- Knowledge cutoff
- 30.06.2025
- 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 | — |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
| Source | Benchmark | Value | Grade | Data retrieved |
|---|---|---|---|---|
| Design Arena | models · 3dwin rate 41.1% · rank 121 | 977Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · codecategorieswin rate 42.6% · rank 123 | 1,029Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · datavizwin rate 49% · rank 111 | 1,068Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · gamedevwin rate 35% · rank 135 | 942Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · uicomponentwin rate 38.4% · rank 126 | 956Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · websitewin rate 43.6% · rank 124 | 1,051Elo | CC — below the B band | 13.11.2026 |
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 | $0.090 | $0.550 | — | — | 96.67% |
| Parasailfp8 | 131,072 | $0.140 | $0.800 | — | — | 100% |
| Venicefp8 | 128,000 | $0.150 | $0.750 | — | — | 100% |
| Nebiusfp8 | 262,144 | $0.200 | $0.600 | — | — | 96.67% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.