"Kimi K3 Is Not Cheap" — Per-Task Cost Lands Near OpenAI's Top Model, ~20× DeepSeek V4
alexinch.com / Hacker News·medium signal
A July 26 post pushes back on the dominant framing of Kimi K3 as a cost breakthrough, arguing that on Artificial Analysis's Intelligence Index it cost only slightly less per task than OpenAI's top model, roughly double GLM-5.2 and about 20× DeepSeek V4. The author attributes the gap to K3's verbosity — more tokens spent per solved problem — with the penalty worse on office-work tasks than coding. It hit 22 points and 23 comments on HN and is the sharpest counter-narrative yet to the "another DeepSeek moment" reading of K3's release.