@giadotai Well, my harness does it for roughly 20+ models, mixing 6-7 of them on every single task given very specific real-time criteria - and I'm assuming I'm far from alone doing this ^^
@Tornatur149828@thsottiaux@NicolasZu Well, yes but we'd be broke :P
I second @NicolasZu 's point: you can sear through your weekly Codex quota in a few hours using Astra - even just Med/high, when Opus 5.5 is virtually limitless now...
@RyanSael@OpenRouter@alexatallah@bcherny Haha! Amazing!Good for you (just followed you as well)
I also got some attention from a couple of cool benchmarks & graphics done with Opus or Pareto! Found it fun :)
OH
MY
GOD.
Opus 5.5, oneshot.
14.0 min
15 api calls
56.6k output tok
2.50M cache reads tok.
113k cache write tok.
$2.54 API equivalent.
DESTROYS GPT6 Astra, Deepseek 4.1, and Pareto 26.9
(Also destroys Fable 5.1 for that matters)