Image created with gemini-2.5-flash-image with claude-sonnet-4-5. Image prompt: Cinematic nighttime field under expansive starry sky, single telescope silhouette in foreground angled downward toward grass, bold white sans-serif text reading DEEPSEEK centered in upper frame against deep navy sky, widescreen composition, film grain, high contrast, minimalist and atmospheric.
You can now run FP8 reinforcement learning on consumer GPUs! Try DeepSeek-R1’s FP8 GRPO at home using only a 5GB GPU. Qwen3-1.7B fits in 5GB VRAM. We collabed with PyTorch to make FP8 RL inference 1.4× faster. Unsloth: 60% less VRAM, 12× longer context. https://x.com/UnslothAI/status/1993358367776186801
Review of Deep Seek OCR | 90/30 Club https://lukeatkins.me/90_30_Club/posts/deepseekocr/
What will the next-gen LLM architecture look like? This question keeps sparking debates — and Zhihu contributor & developer Yuxuan offers a sharp comparison between DeepSeek Sparse Attention (DSA) and Native Sparse Attention (NSA), plus a practical look at implementing DSA https://x.com/ZhihuFrontier/status/1993231992876421156
deepseek-ai/DeepSeek-Math-V2 · Hugging Face https://huggingface.co/deepseek-ai/DeepSeek-Math-V2
deepseek-ai/DeepSeek-Math-V2 https://github.com/deepseek-ai/DeepSeek-Math-V2/tree/main
Leave a Reply