- Chat
Qwen 3.6 35B A3B
Vision-language mixture-of-experts model: 35B total, 3B active parameters.
Qwen model family
About this model
Qwen 3.6 35B A3B is a 35B total, 3B active mixture-of-experts causal language model with a vision encoder. Its model card lists 40 layers, 256 experts with 8 routed and 1 shared expert activated, Gated DeltaNet with Gated Attention, native image and video understanding, a 262,144-token native context, and output guidance of 32,768 tokens for most queries or 81,920 for complex tasks.
- Parameters
- 35B3B active per token
- Native context
- 262,144262,144 tokens served by Arnict.
- Output guidance
- 81,920Published guidance
Capabilities
- Chat completionsUse this model with the chat-completions API when it is marked Available.
- Native input modalitiesText input is listed in the published catalog metadata.
- Long-context input262,144 tokens · 262,144 tokens served by Arnict.
- Published output guidance81,920 tokens · 32,768 for most queries · 81,920 for complex tasks
Published limits
Published model-card facts, kept separate from live availability.
| Metric | Published value | Note |
|---|---|---|
| Arnict API output limit | 81,920 tokens | Maximum output accepted by the Arnict API for this model |
| Parameters | 35B total / 3B active | Upstream model card |
| Context window | 262,144 tokens | 262,144 tokens served by Arnict. |
| Output guidance | 81,920 tokens recommended maximum | 32,768 for most queries · 81,920 for complex tasks |
| Weights / activations | BF16 | Published format |
- 40 layers
- 256 experts; 8 routed + 1 shared activated
- Gated DeltaNet + Gated Attention
- Native image and video understanding
Use qwen/qwen3.6-35b-a3b with an OpenAI-compatible client after this model is enabled.
Related models
Qwen
Qwen 3.8 2.4T A95B
Text-only mixture-of-experts model: 2.4T total, 95B active parameters.
GLM
GLM 5.3 Flash Abliterated
A derivative of GLM 5.3 Flash with modified refusal behavior for general-purpose inference.
GLM
GLM 5.3 Abliterated
A derivative of GLM 5.3 with modified refusal behavior for general-purpose inference.