Qwen / qwen/qwen3.6-35b-a3b

Qwen3.6 35B-A3B - access through LLMTR

Qwen3.6 35B-A3B is Alibaba Cloud's open-weight vision-language Mixture-of-Experts model from the Qwen3.6 family: a sparse architecture that activates roughly 3 billion of its 35 billion parameters per token, delivering much larger model capacity at close to the inference cost of a small dense model. It offers a 262,144-token context window, native tool calling, JSON object mode and JSON Schema structured output. Step-by-step reasoning is ON by default; the reasoning trace is returned separately from the answer, reasoning tokens are billed as output, and reasoning can be switched off per request when it is not needed. It accepts text and image input at variable aspect ratios and resolutions and produces text only; audio and video input are not supported on this row. Tool calling runs at the model's own discretion — tool_choice forms that force a tool call from the request side are not supported for this model. The context window is SHARED between prompt and output: there is no separate output ceiling beneath it, and the largest output you can request shrinks as the prompt grows. No prompt-cache discount is applied; every request bills at the full input rate.