News for hackers. Stuff that matters.
Re:Microsoft AI: Seven new world-class models at scalable pr... by Ampersander on Wednesday November 18, 2026 @12:12PM (#50548955)
The performance is really fuzzy too, but the price you do know for certain.
Re:Microsoft AI: Seven new world-class models at scalable pr... by mcintyre1994 on Wednesday November 18, 2026 @12:31PM (#50549261)
Price is quite fuzzy too though when it's pricing tokens, because a model that's more token efficient but more expensive per token can be cheaper for a given task. And token efficiency is as fuzzy as performance so you're back where you started.
Re:Microsoft AI: Seven new world-class models at scalable pr... by seizethecheese on Wednesday November 18, 2026 @01:23PM (#50550023)
[delayed]
Re:Microsoft AI: Seven new world-class models at scalable pr... by danielmarkbruce on Wednesday November 18, 2026 @01:34PM (#50550189)
Definitely. Cost per task is nice in theory, pretty absurd in practice.
Re:Microsoft AI: Seven new world-class models at scalable pr... by seizethecheese on Wednesday November 18, 2026 @01:41PM (#50550303)
Well, it depends on the data source. Cost per task for benchmarks is absurd. Cost per task for real world usage, where the model identity is randomized (Like Arena or Agent Duel) seems better, though the tasks still are not going to be representative if you're not doing the same stuff as the battle participants.