After the drama, the real question: is OpenAI's 'internal model' already GPT-6.5 pretraining?
willdepue · x · 2026-09-09
OpenAI's Will Depue redirects attention from the settling drama to the more important question: what exactly is this 'internal model' that solved the Millennium Problem? His read: OpenAI likely has (1) already started astra-next / GPT-6.5 pretraining, and (2) it is already beating astra on evals.
Unverified speculation, but it suggests the scale and progress of OpenAI's internal models may far exceed public expectations.
More from Models
- Researcher says he predicted Navier-Stokes would fall first, but not amid bitter controversy — geoffreyirving · 2026-09-09
- 'Context-window AGI': inside a single window, models beat any individual — sir_codes_alot · 2026-09-09
- littmath explains why models cracked Navier-Stokes but Riemann is harder — littmath · 2026-09-09
- OpenAI's Internal Model Cracks Open Math Problems Where Humanity Scored Effectively Zero — stevenheidel · 2026-09-09
- GPT-6 Astra Tops Design Arena's 3D Leaderboard With Elo 1495, 65 Pts Ahead of Kimi K3 — airesearch12 · 2026-09-09
- If the NS proof took humans years, internal models may reach decades-long METR horizons — Realistic_Stomach848 · 2026-09-09