Skip to content

[GitHub Trending] antirez/ds4

8.5 relevance
Score Breakdown
technical depth
9
novelty
9
actionability
8
community
7
strategic
8
personal
9

Scored daily by a customisable AI persona to surface the most relevant engineering leadership news.

Local inference engine for DeepSeek models across GPU backends, highly novel and strategic.

Open Source github.com
DeepSeek 4 Flash and PRO local inference engine for Metal, CUDA and ROCm - antirez/ds4
Summary

Antirez released ds4 (DwarfStar), a narrow inference engine optimized for DeepSeek V4 Flash and GLM 5.2, supporting Metal, CUDA, and ROCm backends. It enables running capable models on consumer hardware like MacBooks with 96GB+ RAM or via SSD streaming, and can turn legacy CUDA servers into multi-user LLM servers using micro-batching. The project is self-contained, not a general GGUF runner, and acknowledges heavy reliance on llama.cpp/GGML while being developed with strong AI assistance from GPT 5.5 and Claude Fable.

Author

antirez