Anthropic releases preview of Model Hardware Standard for computing disclosure
surprisetalk · hn · 2026-08-28
Anthropic has released a research preview of the 'Model Hardware Standard.' This standard aims to unify and standardize the disclosure of computing hardware required for AI model training. The document details how to measure and report compute usage, including chip types, counts, cluster scale, and FLOPs, seeking to increase industry transparency and help researchers better understand the training costs and scale of frontier models.
More from Infra
- Qwen3.8-Flash reportedly costs 1/9th to train vs Qwen3.7-Plus — VraserX · 2026-08-28
- Sovereign AI market hits $1.5T as companies flee US cloud providers — mikeflache · 2026-08-28
- ComfyUI multi-GPU setups: text+VAE on one card, diffusion on the other — hurdurdur7 · 2026-08-28
- Samsung Takes Early Lead in Nvidia’s ‘NVHBM’ with 8-High HBM4E — zephyr_z9 · 2026-08-28
- Dev ships ik_llama.cpp quants for DeepSeek V4 Flash, beats Atomic in paired KLD tests — KeinNiemand · 2026-08-28
- Space data centers are a terrible idea: heat dissipation physics make it prohibitively expensive — SumitGup · 2026-08-28