Qwen3.5 397B A17B
Qwen's 397B-parameter multimodal mixture-of-experts model with 256K-token context for text, image, and video.
Design Arena
Design Arena ranks models on real-world front-end and design tasks — websites, UI components, data viz, SVG and more — through head-to-head human votes, scored as an ELO rating.| Category | ELO | Win rate | Rank |
|---|---|---|---|
| 3D | 1227 | 56.7% | #39 |
| Website | 1221 | 52.6% | #42 |
| Code | 1216 | 52.6% | #41 |
| Data viz | 1208 | 53.2% | #43 |
| UI component | 1208 | 51.4% | #47 |
Model Description
The Qwen3.5 series 397B-A17B native vision-language model is built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency. It delivers...
Author
Get the full context.
Sign up to read complete case studies, access detailed metrics, and unlock all use cases.
Author
Get the full context.
Sign up to read complete case studies, access detailed metrics, and unlock all use cases.