1 post · on APIs, testing and the tools I build with
Writing
Latest
Same rate card, different bill: choosing between Fable 5.1 and Astra
Both models bill $10 per million input tokens. One of them still costs less, and the reason has nothing to do with the price per token.
Read →