Qwen3.5-122B-A10B

The Qwen3.5 122B-A10B 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. In terms of...

Qwen3.5-122B-A10B ranks #192 of 377 models in the default AI Benchmark Hub order (quality first, then cost). At $1.17 blended per million tokens, it is cheaper than 48% of models with published API pricing. Its 262,144-token context window is about the catalog median (262,144 tokens).