Qwen3.5-397B-A17B is the latest visual language model of the Tongyi Qianwen series. It adopts a hybrid expert architecture (MoE) with a total parameter amount of 397B and an activation parameter of 17B. The model natively supports 256K context length, scalable to ~1M tokens. Supports 201 languages and has unified visual language understanding capabilities, tool calling and reasoning thinking modes
Provider: Qwen
Context window: 256K
Pricing: $0.33 input / $2.10 output per 1M tokens
Capabilities: serverless, json_mode, function_calling, structured_output
