Qwen3 235B A22B 2507 by Qwen - a 235B MoE instruction model with 22B active parameters, 262K context, non-thinking output, and stronger long-context reasoning.
Capabilities, design details, and architectural traits
Qwen3 235B A22B 2507 is a non-thinking instruction model in the Qwen3 family. It uses a mixture-of-experts design with 22B active parameters and is built for long-context text and code generation.
| Trait | Documented detail specific to Qwen3 235B A22B 2507 |
|---|---|
| Primary identity | 235B total-parameter mixture-of-experts model with 22B active parameters |
| Operating mode | Supports non-thinking mode and does not generate <think> blocks |
| Core use | Text and code generation with instruction following, reasoning, math, and tool use |
| Model updates | Described as an updated version with stronger long-context understanding and broader knowledge coverage |
| Alignment profile | Improved alignment with user preferences in subjective and open-ended tasks |
The model is defined by sparse MoE activation at a very large total scale. That gives it the capacity of a 235B model while keeping the active compute at 22B per pass.
Its documented behavior is direct instruction output, not a reasoning-trace format. That makes it distinct from thinking variants in the same family.
Independent evaluations · Artificial Analysis
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Alibaba | Anthropic | Anthropic |
| Release Date | July 21, 2025 | July 24, 2026 | June 9, 2026 |
| Knowledge Cutoff | Jun 2025 | May 2026 | - |
| Context & Limits | |||
| Context Window | 262K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | $0.23 Best Input Pricing | $5 | $10 |
| Output Pricing | $0.92 Best Output Pricing | $25 | $50 |
| Modalities | |||
| Inputs | text | textimage | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 18.4 | 63.1 Best Intelligence Index | 62.1 |
| Coding Index | - | 78.0 Best Coding Index | 76.5 |
| Agentic Index | - | 59.2 Best Agentic Index | 56.6 |
Graduate-level reasoning and expert Q&A evaluation.
Extremely difficult logical reasoning and knowledge.
Logical reasoning over long context windows.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
