Qwen3 235B A22B Thinking 2507
Alibaba · Released Jul 2025
A 235B Mixture-of-Experts model from Alibaba that activates 22B parameters per forward pass and supports extended reasoning with a 262K token context window.
- Strengths
- Designed for complex reasoning tasks with enough model capacity to handle demanding inference patterns while maintaining computational efficiency through sparse activation.
- Best for
- Long-form reasoning, code generation, and multi-step problem solving where deep analysis is needed within an extended context.
- Limitations
- The MoE architecture may show less consistent performance on simple tasks where full model capacity isn't needed, and sparse activation patterns could affect latency variability.
Input / 1M
$0.3
Output / 1M
$3.00
Cached input / 1M
—
Context window
131K
Price history
Input (solid)Output (dashed)
Price change
- 30d
- in increased 200.0% out increased 2900.0%
- 90d
- in increased 200.0% out increased 2900.0%
- 1y
- in increased 200.0% out increased 2900.0%
- Since launch
- in increased 200.0% out increased 2900.0%
Snapshots
| Effective | Input | Output | Cached in | Note | Source |
|---|---|---|---|---|---|
| 20 Jul 2026 | $0.3 | $3.00 | — | Imported from OpenRouter | openrouter.ai |
| 1 Jul 2026 | $0.1495 | $1.50 | — | Imported from OpenRouter | openrouter.ai |
| 11 Jun 2026 | $0.1 | $0.1 | $0.1 | Imported from OpenRouter | openrouter.ai |