Qwen/Qwen3-4B-Thinking-2507 · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Qwen3-235B-A22B-Instruct-2507 is a multilingual, instruction-tuned mixture-of-experts language model with 235 billion total parameters, of which 22 billion are active per inference.
The model features significant improvements in instruction following, logical reasoning, math, coding, multilingual coverage, and long-context support up to 262,144 tokens, operating exclusively in non-thinking mode without <think> blocks. It demonstrates strong benchmark results in knowledge, reasoning, coding, and alignment tasks compared to prior versions.