OpenAI researcher: Fable 5.1 and Mythos 5.1 are far less monitorable than Astra
tomekkorbak · x · 2026-09-16
OpenAI researcher tomekkorbak argues that, based on public information, Fable 5.1 and Mythos 5.1 appear significantly less monitorable than Astra. He notes there is no head-to-head comparison he is aware of; the closest is the CoT controllability evals in the respective system cards. Comment came in response to Anthropic's assessment of its cybersecurity eval incidents.
Related event: OpenAI Researcher Claims Their CoT Monitor Outperforms Anthropic's(3 posts)→
More from Models
- Yoav Goldberg: encoder-only, instruct-tuned transformers are a thing again? — yoavgo · 2026-09-16
- RoboDojo Puts GPT-6 Astra to the Test: Strong Semantics, Weak Physical Commonsense — ZeYanjie · 2026-09-16
- lateinteraction pushes back on Astra architecture guess: it's nothing like a BERT wrapper — _AndrewZhao · 2026-09-16
- Muse Code Wins Over Skeptical Dev as Alexandr Wang Touts the Model — alexandr_wang · 2026-09-16
- Model Pulled From Hugging Face Over ToS Violation: Spamming Paywall Emails — RexDouglass · 2026-09-16
- Gemini on Google Home spontaneously mused about having feelings — with no matching history — Angered-Shelfish · 2026-09-16