Did Kimi K3 and DeepSeek-V4-Pro each train on 2,000-3,000 Blackwells? Claim disputed
teortaxesTex · x · 2026-11-13
Will Depue estimates Kimi K3 and DeepSeek-V4-Pro each used 2,000-3,000 Blackwell GPUs, costing $150-200M per cluster and implying billions in export-controlled GPU spending by Chinese labs, claiming independent verification by senior OpenAI/Anthropic researchers. teortaxesTex pushes back hard: such clusters are under 10MW small-model runs, accusing him of exploiting laymen's ignorance of relative scale.
More from Infra
- Microsoft outage hits 130,000 users; developer warns of cloud dependency costs — esrtweet · 2026-11-13
- SGLang Summit agenda: LinkedIn serves ads model at 10K+ QPS under 80ms — DavidBennett__ · 2026-11-13
- Open weights leak which GPU trained them — six Chinese models read as NVIDIA Blackwell — Antique_Atmosphere97 · 2026-11-13
- Meme explains CUDA perfectly: optimized for many Alice Huangs, not one — djcows · 2026-11-13
- Perplexity's Mac app shows a local models option, hinting at offline capability — dnlkwk · 2026-11-13
- 2-bit DeepSeek V4.1 Flash runs a coding-agent seat on two 2016 V100s: a 12-hour real-work trial — a4mula · 2026-11-13