cnygaard/glq

Lattice and trellis-coded quantization for LLM weights — 2–8 bpw, mixed-precision, fused CUDA tensor-core inference on vLLM + HF

2 best points · 1 comments · 2 Hacker News posts · 8 GitHub stars · View on GitHub

Hacker News discussions

• Show HN: Glq LLM quantization using E8 lattice (2 points, 0 comments, by acd, 2026-06-01)
• Glq a port trellis quantization of large language models as a vLLM plugin (2 points, 1 comments, by acd, 2026-09-10)