GPT-6 reportedly trained on just 100K B200/300 GPUs at a single Texas facility
round · x · 2026-09-05
X user somewheresy claims GPT-6 was trained on only about 100K NVIDIA B200/B300 GPUs, all in a single facility in Texas. He expected a far larger cluster, and reads the modest scale as the biggest sign of how early we still are. The claim is unconfirmed.
More from Infra
- Who Needs a GB300? Researcher Makes a Movie on a $500 16GB GPU — francoisfleuret · 2026-09-05
- Minima quantizes all 496 layers of Qwen3.8-27B to NVFP4 W4A4, matching BF16 at 2.9x smaller — pbaylies · 2026-09-05
- 19 latency patterns to cut non-model latency in AI applications — bibryam · 2026-09-05
- Astra reportedly trained on 100,000 GPUs, a staggering compute scale — sudoraohacker · 2026-09-05
- gfx906-llama-cpp Fork Boosts MI50/MI60 Inference +23% Prefill, Fits 250k Context on 40GB — milpster · 2026-09-05
- GeneralistAI collects 500,000+ hours of real robot data, launches onchain data bounties via Robinhood — broodsugar · 2026-09-05