Qwen3 235B-A22B
Qwen's flagship hybrid open model with reasoning and non-reasoning modes and broad multilingual coverage.
Pricing
InputOpen weights
Cached inputn/a
OutputSelf-hosted cost profile
No first-party hosted token pricing on the referenced release post.
Cost bandFree/open
Free tierOpen weights; hosted pricing depends on vendor.
Limits
Context window128k checked Jul 19, 2026
Max outputRuntime-dependent
Knowledge cutoffSee release materials
DeploymentSelf-hosted, Hugging Face, ModelScope, Ollama, vLLM, SGLang
OpenAI-compatible APIYes
ModalitiesText
Where it fits
Best forOpen research stacks, multilingual deployments, and teams that want hybrid think/no-think control.
Business fitA serious open contender when you want frontier-style reasoning behavior without a closed API.
Capabilities
Thinking and non-thinking modes119 languages and dialectsOpen weights
Benchmark scores
Composite83 / 100
Coding87 / 100 (LiveCodeBench)
Reasoning89 / 100 (GPQA Diamond)
Long context78 / 100 (RULER 128K)
Vision68 / 100
Instruction following82 / 100 (IFEval)
Editorial aggregate of published benchmark results as of May 1, 2025. These are not independent measurements run by this site.
Change history
- Jul 19, 2026Context increaseQwen3 235B-A22B context window now 128k
32kto 128k OpenRouter directory - Apr 29, 2025New modelQwen3 235B-A22B open-sourced Hybrid think/no-think modes across 119 languages. Narrows the reasoning gap on open-weight models.
Data
Last checkedJul 19, 2026
Cross-checked againstopenrouter.ai/qwen/qwen3-235b-a22b
Verify with provider documentation before committing spend. Report an error and it gets fixed with a logged correction.