🧮 AI Inference Cost Calculator
Compare inference pricing across US, Chinese, and open-weight models.
Inspired by "Who's Afraid of Chinese Models?"
by Ben Thompson — where COGS meets the reasoning era.
| Model |
Provider |
Input $/M tok |
Output $/M tok |
Total Cost |
💡 How to read this: Prices are per-million-tokens at API rates as of July 2026.
Kimi K3 ($3/$15) costs the same as Fable 5 on paper — but Chinese models aren't just competing on price.
The real story is capabilities per dollar in the reasoning era, where chain-of-thought multiplies output costs.
Open-weight models (DeepSeek, Qwen, GLM) can be self-hosted, making marginal inference cost approach zero for high-volume use.