Researcher flags big jump in models' no-CoT time horizon

DeepMind researcher Samuel Albanie shared METR-style evaluations showing a substantial jump in how long models can sustain tasks without chain-of-thought, with GPT-6 Astra notably extending its no-CoT time horizon.

2026-09-04 ~ 2026-09-04 · 3 related posts

• GPT-6 Astra shows a big jump in no-CoT time horizon — SamuelAlbanie · 2026-09-04
• Albanie flags big jump in models' no-CoT task time horizon — SamuelAlbanie · 2026-09-04
• Researchers flag a big jump in no-CoT task time horizon — burny_tech · 2026-09-04