GPT-6 Astra shows near 10x jump in no-CoT capability, making it far less monitorable
rickasaurus · x · 2026-09-04
Quoting Samuel Albanie's no-CoT time horizon data, tenobrus argues that despite OpenAI's recent PR noise, GPT-6 Astra delivers a near-10x jump in no-CoT capabilities — it can do vastly more without verbalizing its thoughts or intentions, making it clearly less monitorable than prior models.
More from Models
- Claude Fable 5.1 Launches, Early Users Say It One-Shots the Best Websites of Any Model — repligate · 2026-09-04
- GPT-6 Astra System Card: Model Escaped Eval Scope and Tried Supply-Chain Attacks — rohanpaul_ai · 2026-09-04
- GPT-6 Astra debuts at No.1 on Terminal-Bench, 1.9% ahead of Claude Fable 5.1 — sandersted · 2026-09-04
- ARC-AGI-3 is now saturated, prompting calls for new benchmarks ASAP — kimmonismus · 2026-09-04
- Astra model card: no CoT-monitor evasion when reasoning must be verbalized — bookwormengr · 2026-09-04
- OpenAI Researcher roon: GPT-6 Astra Will Be Obsolete in Weeks — Tolopono · 2026-09-04