Qwen3-VL-235B-A22B
by Alibaba
· QWEN3 Vl family · released Sep 25, 2025
Open weights
Active
Alibaba's open-weight flagship vision-language MoE, the first open model to surpass GPT-4o and Claude Sonnet 4 on vision benchmarks like MMMU and OSWorld. It supports image, video, and GUI agent tasks with 256k context.
No pricing recorded for this model yet.
Specifications
| Model type | Multimodal |
|---|---|
| Architecture | MoE |
| Parameters | 235 (22 active) |
| Context window | 262.1K tokens |
| Max output | Not disclosed tokens |
| Knowledge cutoff | Not disclosed |
| License | Apache License 2.0 |
| Modalities | Text input Image input Video input Text output |
| Tags | Agentic Long context Vision |
| End of life | Not disclosed |
| Weights | Download |
| Official page | Link |
| Last verified | Not disclosed |