Mistral API Pricing & Cost Calculator (2026)
Current Mistral API pricing per million tokens — Medium, Small and the open-weight family — with a live cost calculator, worked examples, EU-jurisdiction notes and savings tips.
Mistral is the leading European LLM provider and a favourite where data residency and open weights matter. Its catalogue mixes managed-API models with openly released ones, and its pricing tends to keep input and output closer together than the big US frontier labs — which can make it surprisingly economical for output-heavy work. This page covers the lineup, a worked example, why teams pick Mistral and how to keep the bill down. The table above is from a live feed; the cost calculator shows your real spend.
The Mistral lineup
Mistral's range, billed per million input and output tokens, spans:
- Mistral Medium — the balanced workhorse, competitive with mid-tier models from larger labs, usually at a lower price.
- Mistral Small — fast and inexpensive, for high-volume, simpler tasks.
- Open-weight family (Ministral and others) — compact models you can run via the API or self-host.
- Specialist models — including coding-focused options for code generation and completion.
The table above shows the live spread. A Batch tier offers roughly a 50% discount for asynchronous bulk jobs.
Why teams choose Mistral
- European jurisdiction. For organisations with EU data-governance requirements, a European provider materially simplifies compliance and procurement.
- Open weights. Several Mistral models are openly available, so you can prototype on the managed API and later self-host the same model for full data residency — without re-engineering your prompts. See our self-hosted inference review.
- Output-friendly pricing. Because input and output rates are closer together than at some US labs, response-heavy workloads (chat, generation) can come out cheaper than the headline input price suggests.
- Solid mid-tier price/quality. Mistral Medium is a strong default for general production work that doesn't need the absolute frontier.
A worked example
A chat product averages 1,200 input tokens and 800 output tokens per turn, 80,000 turns a month, on Mistral Medium:
cost_per_turn = (1,200/1e6 × input) + (800/1e6 × output)
monthly = cost_per_turn × 80,000Because output (800 tokens) is a large share of each turn, Mistral's relatively low output price helps here — a workload where a provider with a high output multiplier would cost more. Compare the same scenario against OpenAI and DeepSeek in the calculator to see where Mistral lands for your input/output ratio.
How to cut your Mistral bill
- Default to Small for simple, high-volume tasks; use Medium where quality matters.
- Cap output — even with friendlier output pricing, length still drives cost.
- Batch offline jobs for the asynchronous discount.
- Consider self-hosting an open-weight Mistral model once volume is high and steady, for fixed cost and full data residency.
When it's the right pick
Choose Mistral when European jurisdiction or an open-weight migration path matters, when your workload is output-heavy, or when you want a capable mid-tier model without frontier pricing. For the very top of reasoning quality you'll still compare against OpenAI and Anthropic; for the absolute lowest price, against DeepSeek. The calculator makes that comparison one click away.
Frequently asked questions
Is Mistral GDPR-friendly? As a European company, Mistral is a common choice for EU data-governance requirements, and its open-weight models can be self-hosted within your own infrastructure for maximum control. Always confirm specifics against your own compliance needs.
Can I self-host Mistral models? Several are open-weight, yes — you can run them on your own GPUs. Whether that beats the API on cost depends on utilisation; see the self-hosted cost breakdown.
How does Mistral Medium compare to GPT-5 or Claude Sonnet? It's a capable mid-tier model, often cheaper, and competitive for general work. For the hardest reasoning the frontier labs may still lead — test on your tasks.
Does Mistral offer batch discounts? Yes, a batch tier gives roughly 50% off for asynchronous bulk processing.
Compare Mistral across the full field in the LLM API cost calculator.
Prices are auto-refreshed from a live source and dated. Confirm current pricing on Mistral's page before committing.