NVIDIA lands nvfp4 in large-scale pretraining under codename "yolo bundle"
SonglinYang4 · x · 2026-11-14
SonglinYang reveals NVIDIA has landed nvfp4 in pretraining under the codename "yolo bundle". Researcher jackcookjack had been skeptical nvfp4 pretraining would ever work for actual large-scale runs, but it did—a blog post explains how it can be done. Stable low-precision pretraining at scale has direct implications for training cost and compute economics.
More from Infra
- GAO: 76% of US air traffic control systems unsustainable, still on floppies and Win95 — MartinGTobias · 2026-11-14
- MoE combine traffic cut from 16 to 8.25 bits per value via RHT + S2F6 quantization — snowclipsed · 2026-11-14
- Report claims Jensen Huang resigns as Nvidia CEO; CFO Colette Kress named interim CEO — pdamodaran · 2026-11-14
- Drogue raises $20M to power floating data centers with the ocean's heat — arian_ghashghai · 2026-11-14
- AWS Explains Bedrock AgentCore V2 Snapshot Startup: Sessions Ready in ~2 Seconds — AWS ML Blog · 2026-11-14
- Megaspeed was NVIDIA's largest chip buyer in Southeast Asia by 2025, now official cloud partner — RileyRalmuto · 2026-11-14