Daniel Miessler walks back GPT-6.2 Astra hype: cheap Fable-parity on benchmarks, but worse in use
daniel_mac8 · x · 2026-11-18
Daniel Miessler initially said GPT-6.2 Astra / Sol hits rough parity with Fable 5.2 on several benchmarks at a much lower price — "OpenAI is back" (while noting Ant has Fable 5.5 / Model 4 up its sleeve).
He then walked it back after hands-on testing: on an open-ended question about durable moats for the next decade, he found Fable 5.2 clearly superior in depth and insight, feeling "something off" with Astra 6.2. After only 45 minutes he's not declaring it dead, but real-use impressions are not looking good; he's still testing.
More from Models
- GPT-6.2 Astra tops Mercor's OpenWorldEngineer benchmark at 67.2% — morqon · 2026-11-18
- GPT-6.2 Astra takes top spot on OpenAI ECI despite mixed reviews, chart shows — ChrisGPT · 2026-11-18
- OpenAI Questioned for Shipping GPT-6.2 Astra and Sol Despite Tester Backlash — Angaisb_ · 2026-11-18
- Microsoft's MAI-Image-2.7 exits preview, now #2 lab across all three Image Arenas — arena · 2026-11-18
- Microsoft-Decision-1.1 costs $140 per million Artificial Analysis Decision Index tasks — ArtificialAnlys · 2026-11-18
- GPT-6.2 Astra and Sol land #7 and #9 on Vals Index, Sol costs less than a third per task — rickasaurus · 2026-11-18