qwen
Qwen: Qwen3.5 397B A17B
qwen/qwen3.5-397b-a17b
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
qwen/qwen3.5-397b-a17b is a post-trained Qwen3.5 vision-language model that accepts text, images, and video and produces text. Its model card describes a 397B-parameter hybrid mixture-of-experts design with 17B parameters activated, a native 262,144-token context extensible to about 1.01M, and claimed support for 201 languages and dialects. Vendor documentation reports benchmark coverage across knowledge, instruction following, long context, STEM, and reasoning, while OpenRouter catalogs limited external coding, agentic, and design-arena measurements; these remain distinct from source-reported capabilities.[1][2]
Strengths
- The catalog lists text, image, and video inputs with text output, while the model card describes unified vision-language training.[1][2]
- The hybrid sparse-MoE design totals 397B parameters with 17B activated and provides a native 262,144-token context, extensible by the card to about 1.01M.[2]
- The card documents Apache-2.0 distribution and compatibility with Transformers, vLLM, SGLang, and KTransformers; the catalog also lists tools, tool choice, structured outputs, and response formatting.[1][2]
Limitations
- Evidence is limited to vendor documentation and one catalog entry; no independent workload evaluation, safety assessment, or reliability data is supplied.[1][2]
- Knowledge cutoff and hardware requirements are not provided, and the card attributes certain production features, including a 1M-token default context, built-in tools, and adaptive tool use, to hosted Qwen3.5-Plus rather than this exact variant.[1][2]
- Reasoning is optional in the catalog but no default or supported effort levels are listed; its cataloged agentic measurement is also lower than its coding measurement, so reasoning-heavy agents need direct testing.[1]
Where to use it
- Multimodal assistants that answer in text from mixed text, image, or video inputs, with quality checked on the intended media.[1][2]
- Long-document analysis where the large native context may help; test any extended-context configuration separately.[1][2]
- API assistants needing tool calls, structured responses, or multilingual coverage; validate tool behavior, agentic reliability, and each target language.[1][2]
Sources
- [1] OpenRouter catalog · Qwen: Qwen3.5 397B A17BSources checked: Nov 13, 2026
- [2] Model documentation · Qwen: Qwen3.5 397B A17BSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
The Qwen3.5 series 397B-A17B native vision-language model is built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency. It delivers...
Specifications
- Model ID
- qwen/qwen3.5-397b-a17b
- Canonical ID
- qwen/qwen3.5-397b-a17b-20260216
- Creator
- qwen
- Added to OpenRouter
- 16.02.2026
- Context length
- 262,144
- Modality
- text+image+video->text
- Input modalities
- text, image, video
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.600 / out $3.6
- Knowledge cutoff
- —
- 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 | reasoning: optional · parameters: include_reasoning, reasoning |
| Coding | BB — next 30% of the cohort | — |
| Agents | CC — below the B band | parameters: structured_outputs, tool_choice, tools |
| Writing | No measurement | output: text |
| Reasoning | No measurement | reasoning: optional · parameters: include_reasoning, reasoning |
| Document analysis | No measurement | — |
| Vision | No measurement | input: image |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | input: video |
| Embeddings | No measurement | — |
Evaluations
| Source | Benchmark | Value | Grade | Data retrieved |
|---|---|---|---|---|
| Artificial Analysis | Coding index | 48.2 | BB — next 30% of the cohort | 13.11.2026 |
| Artificial Analysis | Agentic index | 8.3 | CC — below the B band | 13.11.2026 |
| Design Arena | models · 3dwin rate 56.7% · rank 72 | 1,144Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · codecategorieswin rate 52.7% · rank 72 | 1,170Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · datavizwin rate 53.2% · rank 69 | 1,173Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · gamedevwin rate 50.2% · rank 84 | 1,131Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · svgwin rate 56.1% · rank 57 | 1,131Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · uicomponentwin rate 51.4% · rank 83 | 1,154Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · websitewin rate 52.6% · rank 74 | 1,185Elo | BB — next 30% of the cohort | 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) |
|---|---|---|---|---|---|---|
| Alibaba | 262,144 | $0.390 | $2.34 | — | — | 100% |
| DeepInfrafp8 | 262,144 | $0.450 | $3 | — | — | 100% |
| Parasailfp8 | 262,144 | $0.500 | $3.6 | — | — | 100% |
| AtlasCloudfp8 | 262,144 | $0.550 | $3.5 | — | — | 100% |
| StreamLake | 256,000 | $0.600 | $3.6 | — | — | — |
| GMICloudfp8 | 262,144 | $0.600 | $3.6 | — | — | — |
| Novita | 262,144 | $0.600 | $3.6 | — | — | — |
| Venice | 128,000 | $0.750 | $4.5 | — | — | — |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.