Loading the catalogue…
Open-weight Qwen multimodal MoE model with about 3B active parameters, a 262K-token context, optional reasoning and text and image input.
| Inference Provider | Sovereignty | Runs in | In / out per 1M | Status | Pin |
|---|---|---|---|---|---|
| L3 | €0.25 / 1.50 | Online | |||
| L0 | €0.09 / 0.86 | Online |
14 use cases classified by the curator. The note says why.
The family is described as built for agentic coding, and this variant combines tool-calling with extended thinking for coding workflows that require reasoning and actions.
The 262,144-token context window supports questions over long documents, while vision capability enables handling visual document inputs.
Tool-calling is explicitly supported, making the model suitable for applications that invoke structured functions.
Extended-thinking mode is a direct architectural fit for multi-step planning, and tool-calling supports carrying plans into action.
Tool-calling support and extended-thinking mode provide direct capabilities for selecting and coordinating tools across multi-step workflows.
Tool-calling support can drive browser actions, while extended thinking can help sequence them; no browser-specific benchmark is provided.
The Qwen3.6 family is described as built for agentic coding, providing relevant coding signal, but no completion benchmark is available for this variant.
The family description identifies agentic coding as a design focus, and the 262,144-token context can accommodate large code contexts; no code-review benchmark is provided.
Vision capability directly supports image question answering, though no vision benchmark or image-specific performance data is supplied.
The 262,144-token context window supports working with extensive source material and long outputs, though no writing benchmark is supplied.
Vision capability supports interpreting screenshot inputs, but no screenshot-specific benchmark is available.
Sourced numbers, not our own evaluations. Percentiles are against the whole Stav catalogue.
Computed from aa-intelligence-index, gpqa-diamond, aa-engineering-index, aa-coding-index, livecodebench, terminal-bench-v4, ifbench, aa-lcr.
curl https://api.stav.ai/v1/chat/completions \
-H "Authorization: Bearer $STAV_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen/qwen3-6-35b-a3b",
"reasoning_effort": "high",
"messages": [{"role": "user", "content": "Hello!"}]
}'→ X-Stav-Request-Id · X-Stav-Model · X-Stav-Provider → X-Stav-Sovereignty-Level — the rung the request was served at