#QwQ-32B

Alibaba's QwQ-32B Open-Sourced: How 32B Parameters Rival DeepSeek R1's 671B

Alibaba's open-source reasoning model QwQ-32B achieves performance rivaling DeepSeek R1 (671B) with only 32B parameters through a two-stage reinforcement learning strategy on verifiable tasks.