Mistral API: Free Mode, API Keys, Models and Pricing

The Mistral API is free to start. Mistral's quickstart says API access is "enabled by default with no credit card required" in its free mode, with usage and rate limits.^[1] You create a key in the Studio console at console.mistral.ai, then call https://api.mistral.ai/v1/chat/completions with it as a Bearer token.^[1][2] Mistral's flagship, Mistral Large 4, costs $0.70 per million input tokens and $2.10 per million output tokens.^[3]

One detail surprises people: the API is no longer only Mistral models. Its price list also includes Z.ai's GLM 5.3 and DeepSeek V4.1 Flash. This guide covers the key setup, the full model and price list as of November 4, 2026, and how to choose.

How to get a Mistral API key

From Mistral's Studio quickstart:^[1]

1. Sign in to the Studio console at console.mistral.ai.
2. Open API Keys in the left sidebar and click Create new key.
3. Name the key, set an expiration date, and choose a connector access scope: shared connectors only, or private and shared connectors.
4. Copy the key straight away. Mistral warns that "the full key appears only once" and cannot be retrieved after you close the dialog.

Store it as MISTRAL_API_KEY. Mistral's official SDKs install with pip install mistralai or npm install @mistralai/mistralai, and a plain HTTP request works too:^[2]

curl https://api.mistral.ai/v1/chat/completions \
  -H "Authorization: Bearer $MISTRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "mistral-large-latest", "messages": [{"role": "user", "content": "Summarize this in one line: ..."}]}'

Mistral recommends setting expiration dates and rotating keys regularly, which is easy to forget once a key works.^[1]

Mistral API pricing

Per million tokens, from Mistral's pricing page:^[3]

Context and license from Mistral's model selection guide.^[4] Mistral Large 4 has a cached-input rate of $0.07 per million, a tenth of its input price.^[3]

Beyond chat models, the same key reaches other model types:^[3]

• OCR: Mistral OCR 4.1 at $4 per 1,000 pages, or $5 per 1,000 annotated pages; the older OCR 3 at $2 and $3.
• Speech to text: Voxtral Mini Transcribe 3 at $0.0035 per minute, Voxtral Small Transcribe 3 at $0.004, and a realtime version at $0.008.
• Text to speech: Voxtral TTS 3 at $13 per million characters.
• Embeddings: Mistral Embed at $0.10 per million tokens and Codestral Embed at $0.15.
• Moderation: Mistral Moderation 2 at no charge.

The model guide also lists Shieldstral 1.0, a safety model in Mistral's experimental Labs tier under Apache 2.0, and Timestral 1.0 models for time-series forecasting.^[4] Mistral describes Labs models as experimental and available for a limited time, so do not build long-lived production on them.

Mistral API cost for a typical request

A request with 20,000 input tokens and 2,000 output tokens costs 0.02 × $0.70 + 0.002 × $2.10 = $0.0182 on Mistral Large 4, and 0.02 × $0.15 + 0.002 × $0.60 = $0.0042 on Mistral Small 4. These are list-price arithmetic, not measured bills; reasoning output and cache hits change the real number.

For long documents, the table holds a less obvious answer. Four models on Mistral's list have a 1M-token context, and the cheapest of them is not a Mistral model: Z.ai GLM 5.3 Flash at $0.15 input and $0.50 output, against $0.30 and $1.20 for DeepSeek V4.1 Flash and $0.70 and $2.10 for Mistral Large 4.^[3][4] Price is not quality, so run your own documents through each before committing.

What else the platform offers

The API sits inside a wider platform. Mistral's quickstarts cover building an agent that calls external functions, uploading documents to a knowledge Library for retrieval-augmented generation, durable workflows, and its Vibe CLI, a terminal tool that uses the same API key.^[6] If you only need model calls, none of this is required, but it explains why the console asks about connector scopes when you create a key.

Which Mistral model to use

Mistral's selection guide rates each model from 0 to 5 on speed and performance.^[4] In its own scores, Mistral Large 4 gets 5 for speed and 4 for performance with a 1M-token context, while Mistral Small 4 gets 4 and 3.5 with 256K. Codestral is the code specialist at 128K. The Ministral 3 models trade performance (rated 1 to 2) for low, flat prices, which suits classification and short extraction rather than reasoning.

Licensing is a real difference. Mistral Small 4 and the Ministral 3 models are Apache 2.0, so you can also run the weights yourself; Mistral Large 4 uses Mistral's own open-weight license, which you should read before self-hosting it commercially.^[4]

FAQ

Is the Mistral API free?

You can start without a credit card: Mistral's free mode enables API access by default, with usage and rate limits.^[1] Paid usage is billed per token at the prices above.

Is the Mistral API OpenAI-compatible?

Its chat endpoint is /v1/chat/completions with Bearer authentication and a messages array, the same shape as OpenAI's Chat Completions.^[2] Mistral also ships its own SDKs.

Which models on the Mistral API have the longest context?

Four list 1M tokens: Mistral Large 4, Z.ai GLM 5.3, Z.ai GLM 5.3 Flash and DeepSeek V4.1 Flash. Mistral Small 4 and the Ministral 3 models list 256K, and Codestral 128K.^[4]

What does mistral-large-latest point to?

It is the alias Mistral's quickstart uses for its large model; check the model card for the version the alias currently resolves to before pinning production code.^[2]

Mistral API or another provider

If you want Mistral's own models, especially Mistral Large 4 or Codestral, Mistral's platform is the direct route and free mode makes testing cheap. For the third-party models it hosts, compare prices before choosing: Mistral lists DeepSeek V4.1 Flash at $0.30 input and $1.20 output, below reAPI's DeepSeek V4.1 Flash at $0.40 and $1.30, checked November 4, 2026.^[5] reAPI does not serve Mistral's own models.

References

1. Mistral AI. Activate Studio and generate an API key. Retrieved November 4, 2026 from docs.mistral.ai
2. Mistral AI. Send your first API request. Retrieved November 4, 2026 from docs.mistral.ai
3. Mistral AI. API pricing. Retrieved November 4, 2026 from docs.mistral.ai/inference/pricing
4. Mistral AI. Model selection guide. Retrieved November 4, 2026 from docs.mistral.ai/inference/model-selection-guide
5. reAPI. DeepSeek V4.1 Flash model page. Retrieved November 4, 2026 from reapi.ai/models/deepseek-v4-1-flash
6. Mistral AI. Documentation index (quickstarts and Labs). Retrieved November 4, 2026 from docs.mistral.ai/llms.txt