Currently listed through these providers:
Model details
Qwen 3.6 35B A3B Uncensored Thinking
Qwen 3.6 35B A3B Uncensored Thinking is a Mixture-of-Experts variant in the Qwen 3.6 family that activates a smaller subset of parameters per token while keeping the broader 35B expert pool available for harder calls. The "A3B" sizing in the model key indicates an active parameter count in the single-digit billions, a configuration that aims to balance reasoning depth with throughput efficiency. It is offered with open weights and a thinking mode enabled, giving the model room to deliberate step by step before producing a final answer. The checked capabilities line up with that intent, covering text and image inputs, attachments, structured output, tool calling, and explicit reasoning support.
In practical terms, this release is shaped around deliberate coding, multimodal analysis, tool use, and complex conversational tasks, all of which benefit from the extended thinking phase. The large 262.1K token context window paired with a 32.8K maximum output makes it suitable for long-document reasoning, multi-step agentic loops, and projects that combine images with extended code or writing. Quantized NVFP4 serving through Aoru AI delivers higher throughput than the auto-routed endpoints while keeping cache reads inexpensive, which helps when repeatedly revisiting the same long prompts during iterative development. Overall, the model fits well with users who want open-weight flexibility, strong reasoning control, and efficient serving for agent-style workloads.
Quick Info
Powered by- Provider
- NanoGPT
- Model key
- qwen/qwen3.6-35b-a3b-uncensored:thinking
- Release date
- Jul 29, 2026
- Last updated
- Aug 24, 2026
- Input modalities
- Output modalities
- Capabilities
Cost
- Input token cost
- $0.15
- Output token cost
- $0.95
Limits
- Input tokens
- 262,144 tokens
- Output tokens
- 32,768 tokens
- Context window
- 262,144 tokens