Ingot network quantizes Qwen 27B to FP8 better than Alibaba's own in one week
const_reborn · x · 2026-10-22
constreborn argues "all models will be trained in markets," highlighting the newly launched model network Ingot (97):
- Its latest release quantized Qwen 27B to FP8 with better quality than Alibaba's own quantization, in just one week
- Next targets: 2-bit quantization and the larger GLM models
- The thesis: distributed, market-based networks for model training and optimization
More from Infra
- Eden Deploys Vercel's Eve Agent to Cloudflare With One Command, Fully Open Source — ritakozlov · 2026-10-22
- Node.js ships emergency fix for high-severity Brotli decoder RCE in v22/24/26 — cramforce · 2026-10-22
- PyTorch's TorchTitan adds DeepSeek-V4, Kimi K3 support, RL training and big scale gains — PyTorch · 2026-10-22
- Epoch AI tracks 95 AI data centers with 14.2 GW — still just 44% of global AI compute — haider1 · 2026-10-22
- Two CVSS 9.2 RCE Vulnerabilities in PHP Push a European Government Site Offline — HankYeomans · 2026-10-22
- Old H100s Now Rent Above Launch Prices as Compute-per-Watt Value Keeps Climbing — abhiadesai · 2026-10-22