Learn

Plain-English guides to Mistral AI, La Plateforme, and the broader open-model ecosystem. No fluff, no affiliate spam.

Mistral basics

Mistral's pricing is per million tokens, split across input and output, with real discounts on batch and embed. Here's how to model it.

Both models sit on La Plateforme. One is a frontier flagship; the other is 5–10× cheaper. Here's when each actually wins.

Function calling on Mistral is OpenAI-compatible in shape, but with stricter JSON-schema validation. Here's the working pattern.

Codestral is the model to reach for on autocomplete, refactor, and tight code generation. It's not the right choice for full agentic coding.

A focused glossary of the terms that show up in the Mistral docs, dashboards, and error messages — plus a few from the surrounding LLM ecosystem.