Early user hands-on: OpenAI's Astra model is noticeably faster than Fable

tristanbob · x · 2026-09-05

User tristanbob compared Astra with its predecessor Fable, reporting that Astra is very fast — whether from higher tokens per second or better token efficiency, he isn't sure. He jokes it should have been called "Fastra". Early hands-on feedback on OpenAI's unreleased model.

Related event: GPT-6 Astra early reviews: record ARC-AGI scores and big speed gains, but benchmark gaps spark debate(16 posts)→

Original post →

More from Models

Models channel →