Qwen: Qwen3.5 397B A17B

The Qwen3.5 series 397B-A17B native vision-language model is engineered with a cutting-edge hybrid architecture, integrating a linear attention mechanism with a sparse mixture-of-experts model. This design significantly boosts inference efficiency while delivering state-of-the-art performance across a broad spectrum of tasks. Users can leverage its capabilities for advanced language understanding, complex logical reasoning, robust code generation, and sophisticated agent-based applications. This model stands out for its exceptional performance in image and video understanding, as well as graphical user interface (GUI) interactions. With a generous context window of 262K tokens and a max output of 65K tokens, it supports extensive and detailed interactions. Key capabilities include JSON mode, streaming, functions, long context, video input, structured output, and vision. Access Qwen3.5 397B A17B on Multi AI for just $0.15/1.00 per 1M tokens (input/output).