Qwen3: Think deeper, act faster

Qwen3, the newest in the Qwen large language model family, has been released. The flagship Qwen3-235B-A22B model shows competitive performance in coding, math, and general tasks. Eight models, including two MoE and six dense, are now open-weighted under Apache 2.0. Qwen3 offers hybrid thinking modes, multilingual support for 119 languages, and improved coding and agentic capabilities. It's easily integrable into various platforms, aiming to advance global research in large foundation models.