qwen/qwen3.5-397b-a17bThe 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.
Scores on standardized evaluations. Higher is better — and rank shows where Qwen3.5 397B-A17B lands among all models tracked on Model Beat.
No standardized benchmark scores from Epoch AI yet — common for very recent, image, or niche models. Scores appear here once published.
Model & benchmark data from Epoch AI (CC BY); pricing, specs & descriptions from OpenRouter.