UK AISI assessment: Astra reasons in a more compressed style than GPT 5.6 Sol

maksym_andr · x · 2026-09-04

An independent assessment from the UK AI Security Institute notes that Astra reasons in a compressed style, to a greater degree than GPT 5.6 Sol or GPT 5.5. Its raw reasoning is generally understandable, though phrases with unclear meaning appear more frequently.

Related event: UK AISI Evaluation Finds GPT-6 Astra Solves 30-Minute Human Tasks Without CoT(5 posts)→

Original post →

More from Models

Models channel →