EvoPress quantization shrinks 30B model from 56GB to 19GB at near-bf16 accuracy
TimDarcet · x · 2026-10-21
TimDarcet details EvoPress, a quantization scheme found via evolutionary search: average accuracy 75.6% vs 75.7% for bf16, with average response length 11417 vs 11070 — essentially lossless while cutting weights from 56GB to 19GB. Comparisons: Greedy-KL 75.3%, Q4KM 74.8%, NVFP4/MXFP8 74.4%.
Related event: Meta Engineers Train a 30B LLM from Scratch and Squeeze It onto a 32GB GPU(4 posts)→
More from Models
- X trends: SophontAI's Holon-S1 predicts gene activity from cancer slides; decision models rise — iScienceLuvr · 2026-10-21
- OpenAI's MultiLangBench: PHP scores highest (91.3%) at building a database engine from scratch — ricklamers · 2026-10-21
- 704-vote robotics Elo ranks Claude Opus 5.5 first at 1379; no model completes the task — qinzytech · 2026-10-21
- DeepSeek V4 called 'god-tier infra, 4th-tier ML': 7 tricks stacked, 20 patches to fix them — teortaxesTex · 2026-10-21
- GPT-6 Astra and Claude Fable 5.1 write $10,000 research papers — experts reject all — sayashk · 2026-10-21
- GitHub Copilot's HydraFusion Orchestrates Multiple Models Per Task — WirelessLife · 2026-10-21