Mistral how to get an API key: billing, steps and free tier

Mistral API key, how to get one: sign in to Mistral's Studio console, make sure billing is active, open the API Keys page and select New key. Mistral's docs now say that creating API keys and using the API require billing activation and pay-as-you-go, and that the free mode does not include API access. That differs from older guides, which describe a free Experiment plan.

This guide is for developers who want a working Mistral key and a first request. Everything below comes from Mistral's docs, read on November 8, 2026. The console itself sits behind a login, so we could not open it, and the exact button labels may differ a little from the docs.

If you are comparing vendors, our guides to getting a Claude API key, a Gemini API key and a DeepSeek API key follow the same path.

Is there a free Mistral API key? The Experiment plan question

As of November 8, 2026, Mistral's docs do not mention an Experiment plan. The page on usage and limits says this instead:

• Free mode includes a monthly usage budget for trying Vibe and Studio, Mistral's coding agent and its build console. It does not include API access.
• "Creating API keys and using the API require billing activation and pay-as-you-go."
• Pay-as-you-go covers usage beyond the included monthly amount and draws from prepaid credits. It is not a separate API plan.

The workspace quickstart agrees: "API keys are inactive until a payment method is on file." So a key you create without billing will not work.

Many guides online still describe an Experiment plan with a free API tier that needs only a verified phone number. We found those claims in third-party posts and could not confirm them in the current docs. If you see an Experiment option on your Subscriptions page, that page is the authority for what your account gets. Otherwise, plan on adding a payment method.

What you need first

• A Mistral account that can open the Studio console at console.mistral.ai.
• A payment method. Billing is set up in the Admin Panel under Subscriptions, then Billing. You can connect a credit card or a cloud billing account (AWS or Azure Marketplace), and optionally set a monthly spending limit before saving.
• Patience of two to three minutes. The docs say billing takes that long to take effect, and team members who see inactive keys right after setup should wait and retry.

How to get a Mistral API key

The steps follow Mistral's API key quickstart.

1. Activate billing as described above, in the Admin Panel under Subscriptions, then Billing.
2. In the Studio console, choose the workspace the key should belong to. Mistral says "Each workspace has its own API keys", and usage is tracked and billed per workspace.
3. Open the API Keys page and select New key.
4. Give the key a name (the docs use "quickstart"), set an expiration date, and choose its connector access: shared connectors only, or private and shared connectors.
5. Copy the key right away. The full key is shown only once, so put it in a password manager or secrets vault.

Mistral recommends an expiration date on every key and regular rotation: create a new key, switch your code over, then delete the old one. Delete any key you no longer use or think has leaked.

How to store and test the key

The Mistral SDKs and the docs examples read the key from the MISTRAL_API_KEY environment variable.

export MISTRAL_API_KEY="your-api-key"

That lasts only for the current terminal session. Then send a test request to the chat completions endpoint, as the quickstart does:

curl https://api.mistral.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $MISTRAL_API_KEY" \
  -d '{
    "model": "mistral-small-latest",
    "messages": [{"role": "user", "content": "Hello, Mistral!"}]
  }'

A good response has the reply in choices[0].message.content, plus id, created and usage fields. We did not run this request, because it needs a funded account.

How to make a first request with Python

Mistral's first request quickstart needs Python 3.10 or newer and uv, and installs the mistralai package. Its sample, lightly adapted:

import os
from mistralai.client import Mistral

client = Mistral(api_key=os.environ["MISTRAL_API_KEY"])

reply = client.chat.complete(
    model="mistral-small-latest",
    messages=[{"role": "user", "content": "In one sentence, what is an API key?"}],
)
print(reply.choices[0].message.content)

Run it with uv run <filename>.py. The docs also show TypeScript and curl tabs.

Mistral API limits and spending caps

Mistral does not publish a fixed rate-limit table in these docs. Completion limits are listed per model in the Admin Panel under API, then Limits, as two numbers: tokens per minute and requests per second. Audio and document OCR limits are listed there too. Third-party sites quote round figures for a free tier, but we could not match them to Mistral's own pages, so we do not repeat them.

You can cap spending in two places:

• The organization monthly cap applies to API and Vibe use. Workspace caps cannot exceed it. If you reach it, API access can be suspended until the next month or until an admin raises the limit.
• A workspace budget, set on the billing page. When the workspace hits it, requests return 429 Too Many Requests until the next billing cycle.

To raise limits, contact support at help.mistral.ai with the models, your target requests per second, expected tokens per minute and per month, and a short description of the use case.

Common problems

• 401 with "Invalid API Key": the key is missing, incomplete or revoked. Export the full key again in the same terminal where you run the code. If the key is new, also check that billing is active, since keys are inactive without a payment method.
• 429 Too Many Requests: you hit a rate limit or a workspace spending cap. Wait and retry with exponential backoff, or check the caps above.

What to do next

Check the Subscriptions page for what your plan includes, and Mistral's pricing page for per-model prices, which we did not verify for this guide. Set a workspace budget before you point real traffic at the key.