Astra found less CoT-monitorable, up to 10x better without reasoning chains
birchlse · x · 2026-09-04
Researcher xuanalogue shared an analysis of the Astra model with three findings:
- Astra is in fact less monitorable via chain-of-thought (CoT)
- It can perform up to 10x better without CoT
- Its effective depth is only up to 2x that of previous models
She stresses the concern is not architecture changes or recurrent depth, but that we don't really know why. She adds she would retract the analysis if Astra turned out to run 100 forward-pass equivalents before emitting a token, though trainability issues make that unlikely.
More from Models
- GPT-6 Astra reportedly scores 100% on ExploitBench, finds two zero-days in testing — VraserX · 2026-09-04
- AI launch playbook under fire: influencer hype chorus vs paying users locked out — xeophon · 2026-09-04
- Team finetunes Gemma 12B for audio proofreading, benchmarks it against Gemini — ojasvi_yadav · 2026-09-04
- First impressions of Astra: clean tone and zero jargon in its writing — soumitrashukla9 · 2026-09-04
- Nadella says early customers already use Astra on Azure as Altman responds — i_dg23 · 2026-09-04
- Abu Dhabi institute IFM releases 6 fully open-source AI models with data, code & methods — Polymarket · 2026-09-04