M
Provider
Mistral AI API pricing
Mistral publishes a range that spans more than an order of magnitude, from 0.1 per million input tokens on the smallest lightweight model to 1.5 on the medium tier, and from 0.1 to 7.5 on output, with one experimental entry priced at 0 throughout. Two shapes coexist: the lightweight line charges the same rate in both directions, while the small, large, medium and code models bill output at three to five times input. Cached input is discounted 90 percent across the board, and cache-write prices are not published separately. Four models carry no published price at all, so no rate is shown for them.
Model families
Several models share one commercial name. These pages compare the members of a family side by side.