The evidence goes against this.
“Training GPT-6 Astra on about 100,000 chips amounted to roughly a billion-dollar training run over about two months.”
Say instead: The passage cites 100,000 chips, a roughly billion-dollar run over about 2 months, but names no model.
2 of 4 review checks done · 2 counted
Contradicted · 0.1520 under ledger-default@0.3.0 at snapshot 27745 · review checks 2 of 4 · provisional until audited. The number is model-conditional, not a share of truth.
Checked in Galedra: score 0.15, mostly against · not yet reviewed by anyone else
“Training GPT-6 Astra on about 100,000 chips amounted to roughly a billion-dollar training run over about two months.”
https://galedra.org/claims/4c5f8598-33e4-8c60-ba5d-d4149e643708/card
Galedra
