Resources // LLM Tokens
Mistral (La Plateforme / AI Studio) free tier for developers
Offer Free tier
Suits for Non-VC-backedVC-backed
Updated Jun 2026
The offer
- Free "Experiment" plan on Mistral AI Studio (formerly "La Plateforme"): API access to Mistral's models for testing, evaluation and prototyping
- No credit card required
- Restrictive rate limits (fine for prototyping, not production); upgrade to the paid "Scale" plan for higher limits, pay-per-use billing, frontier-model access and agents
Who qualifies
- Open to anyone: activate the free Experiment plan
- Requires a verified phone number (one number per plan; no card)
- Heads up: API requests on the free Experiment plan may be used to train Mistral's models. You can opt out in the Admin Console
Community Insights
Developers like the free tier’s large monthly token allowance and global access (not US-only), covering Mistral’s full model lineup including Codestral. The two recurring caveats: the ~1 request/second throttle makes it a prototyping tier rather than a production one, and, newer and important, the free Experiment plan may use your API requests to train Mistral’s models unless you opt out. Paid Scale plans are opted out by default. Note the platform was rebranded from “La Plateforme” to “Mistral AI Studio.”
Best Practices (from community tips)
- Opt out of training if your inputs are sensitive. The free Experiment plan is opted in by default. Toggle it off in the Admin Console (Privacy) before sending anything you wouldn’t want used for training.
- Treat it as a prototyping tier. The ~1 req/sec limit kills concurrency and batch jobs; move to the paid Scale plan when you need real throughput.
- Don’t trust the quota readout instantly. Usage reporting can lag (it refreshes a few times a day), so leave headroom.
- Use it to evaluate models before you spend. Compare Small / Large / Codestral on your own task; the free key works with Mistral’s tooling (Vibe / Codestral CLI).