Run Qwen3.6 35B A3B in the EU through an OpenAI-compatible API

Qwen3.6 35B A3B was the first open-weight Qwen3.6 model, released in April 2026 under Apache 2.0. It has 35 billion parameters in total and about 3 billion active for each token, with a vision encoder for image input. Its 40 layers combine Gated DeltaNet linear attention and gated attention, each followed by a mixture-of-experts block.

The native context window is 262,144 tokens, extendable to about one million according to Qwen. Like the dense Qwen3.6 27B, it focuses on agentic coding, front-end work and repository-level reasoning, and supports thinking preservation, which keeps reasoning from earlier turns available in long sessions.

With only about 3 billion active parameters, this model answers fast and handles high request volumes efficiently, while still offering image input, tool calls and a long context. It works well for coding autocomplete-style help, quick agent steps, triage of screenshots and bulk text processing where a dense model would be slower.

The small active size has a cost: on harder reasoning, long code changes and nuanced writing, the dense Qwen3.6 27B or Qwen3.8 27B give better answers. Qwen advises keeping at least 128K tokens of context free for the thinking mode. For production-critical outputs, compare it against a larger model on a sample of real requests.

inference.eu uses the OpenAI request format, so the OpenAI SDKs, LangChain, LlamaIndex and coding tools work unchanged. Set the base URL to the inference.eu API, use your inference.eu key and set the model to qwen3.6-35b-a3b. The Anthropic Messages format is also available.

Thinking is on by default; pass enable_thinking: false in chat_template_kwargs through extra_body for direct answers, which suits high-volume tasks. Pass preserve_thinking: true to keep earlier reasoning in multi-turn sessions. Send images as image_url content parts. Because the request format is shared across models, switching to Qwen3.6 27B means changing only the model string.

On inference.eu, Qwen3.6 35B A3B runs on infrastructure in the European Union. Prompts, images, code and responses are processed in EU data centres, are not sent to Alibaba in China, and are never used to train any model.

Every customer receives a data processing agreement under Article 28 of the GDPR and our list of subprocessors, so your records of processing can name every party involved. Billing is in euros, on one invoice across all models. The Apache 2.0 licence allows commercial use without user limits or revenue thresholds.