DeepSeek’s distilled new R1 AI model can run on a single GPU

DeepSeek released a smaller version of its R1 AI model, DeepSeek-R1-0528-Qwen3-8B, built on Alibaba's Qwen3-8B. This distilled model outperforms Google's Gemini 2.5 Flash on certain math benchmarks and nearly matches Microsoft's Phi 4 on others. While less capable than full-sized models, it requires significantly less computational power, making it suitable for both research and industrial applications using smaller-scale models.