Qwen3 235B A22B Instruct 2507

Qwen3-235B-A22B-Instruct-2507 is a multilingual, instruction-tuned mixture-of-experts language model based on the Qwen3-235B architecture, with 22B active parameters per forward pass. It is optimized for general-purpose text generation, including instruction following,...

Qwen3 235B A22B Instruct 2507 ranks #94 of 377 models in the default AI Benchmark Hub order (quality first, then cost). At $0.32 blended per million tokens, it is cheaper than 75% of models with published API pricing. Its 262,144-token context window is about the catalog median (262,144 tokens).