Analysis: US Models Are Now Cheaper per Task Than Chinese Ones, Upending a Widely Held Assumption
sebkrier · x · 2026-10-06
Citing Artificial Analysis data, Chorzempa Martin argues two widely held AI assumptions are outdated: Chinese models are not actually cheaper — at almost every intelligence level, US models cost less per task, since they use tokens more efficiently even if priced higher per token, and OpenAI/Anthropic/Google likely have unseen optimizations plus better chips. The caveat: you can't undercut free — open-weight Chinese models can be downloaded with no lab fees, though without own compute, self-hosting savings are shrinking.
More from Venture
- Minerva, a robot startup for extreme jobs, comes out of stealth — kyliebytes · 2026-10-07
- Flai raises $27M Series A led by Base10 to build AI-native CRM for car dealerships — ycombinator · 2026-10-07
- Nvidia nears $6 trillion market cap as 'leather jacket selling thinking rocks' hits capitalist endgame — Kyrannio · 2026-10-07
- Ramp: Record 8% of firms switched top AI model provider in September — annbordetsky · 2026-10-07
- Exa raises Series C from Felicis, betting search must be rebuilt for agents — pxd · 2026-10-07
- Firecrawl launches Alexandria fantasy football MCP data, announces $75M Series B — devdigest · 2026-10-06