Qwen

Qwen3 turns reasoning into a knob instead of a shrine

Qwen3 open-weights a full range of dense and MoE models under Apache 2.0, with hybrid thinking modes that let builders trade speed for deeper reasoning when the task actually deserves it.