HN · GitHub Signals

OPEN SOURCE, IN FOCUS

glm-5.2-4090

renning22/glm-5.2-4090

First known correct serving of GLM-5.2 (DeepSeek-Sparse-Attention) on consumer RTX 4090 GPUs — a portable Ada/sm_89 port of the DSA kernel stack.

Explore on GitHub ↗

Saved projects stay in this browser. No account needed.

TAKE A CLOSER LOOK

Start with the source.

The repository is the place to check installation steps, supported platforms, the license, and current maintenance before choosing a tool.

Read the documentation on GitHub ↗

Keep exploring

Find projects in the same areas, or search the archive for a different approach.

ANOTHER GOOD FIND

Explore related projects

View the archive ↗

Connected by topics and descriptions.

ai-infos

vllm-gfx906-mobydick

A high-throughput and memory-efficient inference and serving engine for LLMs - Optimized for AMD gfx906 GPUs, e.g. Radeon VII / MI50 / MI60

Take a look →
dogacel

auto-gpu-kernel

Winner 🏆 (Agent-only) MLSys 2026 - FlashInfer AI Kernel Generation Contest for the DeepSeek Sparse Attention (DSA) track with an average speedup of 34.93x

Take a look →
giannisanni

pulsar

SSD-streaming inference engine for giant MoE models (Rust + CUDA). GLM 5.2 743B at 2 tok/s and Hy3 295B at 7 tok/s on two consumer 16GB GPUs. Zero-config multi-GPU: measures PCIe bandwidth, places attention and hot experts where they fit.

Take a look →
alibaba

atrex-bench

End-to-end benchmark for AI-generated GPU kernels, drawn from real production traces — turn a PyTorch reference into a DSL kernel (Triton, Gluon, FlyDSL, CuteDSL) and grade it on compilation, numerical correctness, and speed-of-light efficiency, on both AMD and NVIDIA GPUs.

Take a look →