qwen
Qwen: Qwen3.5-35B-A3B
qwen/qwen3.5-35b-a3b
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
qwen/qwen3.5-35b-a3b is Qwen's post-trained vision-language model with 35B total and 3B activated parameters, using a hybrid architecture that mixes Gated DeltaNet linear attention, gated attention, and a sparse mixture-of-experts design for inference efficiency, according to its model card. It accepts text, image, and video input and produces text, with a native context length of 262,144 tokens that the card says can be extended toward roughly one million. The card claims support for 201 languages and dialects. Weights are distributed under Apache-2.0 and are described as compatible with Transformers, vLLM, SGLang, and KTransformers.[1][2]
Strengths
- Sparse activation of roughly 3B of 35B parameters via a hybrid linear-attention and mixture-of-experts design, which the source card positions as delivering high-throughput, low-latency inference.[2]
- Native multimodal input covering text, image, and video per the OpenRouter catalog, paired with a 262,144-token native context that the model card reports as extensible to about 1,010,000 tokens.[1][2]
- Broad deployment surface: open Apache-2.0 weights, stated compatibility with Transformers, vLLM, SGLang, and KTransformers, plus API support for tools, structured outputs, and optional reasoning controls.[1][2]
Limitations
- Reported benchmark results come from the vendor's own tables and have not been independently verified in this source pack; OpenRouter's third-party coding index exists but its agentic and intelligence indices are unpopulated.[1][2]
- Knowledge cutoff, hardware requirements, and detailed video-handling behavior are not specified in the supplied evidence, so real-world performance on specific workloads should be validated before adoption.[1][2]
Where to use it
- Long-context document or multimodal assistants where a 262K-class context window and image/video ingestion are useful, pending workload validation.[1][2]
- Self-hosted or custom-served deployments via vLLM or SGLang for teams wanting an open-weight, tool-calling model with cost-conscious sparse inference.[1][2]
Sources
- [1] OpenRouter catalog · Qwen: Qwen3.5-35B-A3BSources checked: Nov 13, 2026
- [2] Model documentation · Qwen: Qwen3.5-35B-A3BSources 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 35B-A3B is a native vision-language model designed with a hybrid architecture that integrates linear attention mechanisms and a sparse mixture-of-experts model, achieving higher inference efficiency. Its overall...
Specifications
- Model ID
- qwen/qwen3.5-35b-a3b
- Canonical ID
- qwen/qwen3.5-35b-a3b-20260224
- Creator
- qwen
- Added to OpenRouter
- 25.02.2026
- Context length
- 262,144
- Modality
- text+image+video->text
- Input modalities
- text, image, video
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.140 / out $1
- 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 | CC — below the B band | — |
| Agents | No measurement | 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 | 37 | 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) |
|---|---|---|---|---|---|---|
| Darkbloomfp4 | 262,144 | $0.065 | $0.750 | — | — | 100% |
| DeepInfrafp8 | 262,144 | $0.140 | $1 | — | — | 100% |
| Parasailfp8 | 262,144 | $0.150 | $1 | — | — | 100% |
| Venice | 256,000 | $0.150 | $1 | — | — | 100% |
| Alibaba | 262,144 | $0.163 | $1.3 | — | — | 100% |
| AtlasCloudfp8 | 262,144 | $0.225 | $1.8 | — | — | — |
| SiliconFlowfp8 | 262,144 | $0.240 | $1.8 | — | — | — |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.