AI Resonance · 阅读最新日报 · AI 入门推荐

2026-11-13 · America/Los_Angeles · 社区动态 · #7

Weight Space Programming for Quantization: Nace quants achieve up to 2x better KL divergence for Qwen 9B, 27B, 35B & Flash Next [GGUF]

hey llamas, its Alish from nace.ai (open source ai lab behind #1 decision model on latest Artificial Analysis decision index v2.1) & Kapitsa (optimized inference for apple silicon). As promised in the last post, we are releasing the GGUF weights based on the Nace quantization of Qwen 9B, 27B, 35B and Flash-Next. I will describe our process on how we came up with these quants right below, so skip to the end to see the benchmarks and download the weights if it gets too dull. The process I started with 9B quants. I understood that setting up a good & fair benching is critical to succeed with…

Weight Space Programming for Quantization: Nace quants achieve up to 2x better KL divergence for Qwen 9B, 27B, 35B & Flash Next [GGUF]

热度 47.2 / 100;排名与评分保留该期记录。