DeepSeek R1 Distill Qwen 32B

DeepSeek releases distilled 32B model using R1 reasoning outputs applied to Qwen 2.5 32B. Claims state-of-the-art performance for dense models at $0.29/M tokens.