Mistral API Free Tier (Free Mode, formerly Experiment): No Card Needed (2026)

Short answer: yes, Mistral has a free API tier. Every new Mistral account starts in Free mode, which lets you create API keys and call Mistral models without a credit card. You get an included monthly usage allowance plus per-model rate limits. Mistral does not publish the exact numbers; you see them on the Limits page in your own console. Once the allowance runs out, you either wait for the next month or switch on prepaid pay-as-you-go credits.

Checked against Mistral's official docs, help center and pricing page on 2026-10-17. Mistral changed its plan names in 2026, so older guides that talk about the Experiment and Scale plans describe the previous setup.

What you get in Free mode

Mistral's Studio quickstart says: API access is enabled by default in Free mode with no credit card required, and usage and rate limits apply. The subscriptions page adds that Free mode is the default state for new accounts, and that each plan's included monthly usage is shared across Studio, the API and Vibe Code (Mistral's coding agent).

• API keys: you can create keys in Mistral Studio straight away, with your own expiration date.
• Included monthly usage: a usage allowance that resets each billing period. Mistral does not publish a token or dollar figure for Free mode.
• Per-model rate limits: for chat and completion models, limits are set in tokens per minute and requests per second. Audio models can have audio seconds per minute and per month; OCR can have pages per minute.
• Studio access: the pricing page says Free plan users also get access to Mistral Studio, the developer console where you manage keys, usage and billing.
• No card: you do not need a credit card to make API calls in Free mode.

What you do not get

• No published quota. Mistral's docs never give a fixed token count for Free mode. Third-party sites still quote numbers such as 1 request per second or 1 billion tokens per month, but none of these appear on an official page. Treat them as unconfirmed.
• No production guarantees. Free mode is meant for trying models and building prototypes. For more capacity, Mistral's usage and limits page says to contact support with your target requests per second, expected tokens per minute and per month, and a short description of your use case.
• No free overage. When the included usage is used up and pay-as-you-go is off, requests can stop until the next billing period.

Who can use it

Anyone who can create a Mistral account and sign in to Mistral Studio. There is no student, startup or company check for Free mode, and the docs do not list any country restrictions beyond the general availability of Mistral's services. Access works at the organization level: your plan, included usage, spending limit and rate limits apply to the whole organization, not to each person in it.

• Existing customers: if your organization already has a paid plan (Pro, Education, Team or Enterprise), the included usage comes from that plan instead of Free mode.
• Teams: an admin controls the spending limit and the pay-as-you-go setting. If the API stops working for one teammate, it has usually stopped for the whole organization.
• Phone verification: older reports say the Experiment plan asked for a phone number. The current quickstart does not mention it, so it may or may not appear when you sign up.

How to get a free Mistral API key, step by step

1. Go to console.mistral.ai and sign in or create an account. New accounts start in Free mode.
2. In the sidebar, open API Keys and select Create new key.
3. Give the key a name and set an expiration date. Pick a connector access scope: Shared connectors only or Private and shared connectors. This only matters if you use Mistral's connectors; for plain chat completions, either option works.
4. Copy the key right away and store it somewhere safe. Mistral says the full key appears only once and cannot be retrieved after you close the dialog.
5. Open the Limits page in the Admin panel to see the models available to your organization and the tokens-per-minute and requests-per-second limits for each.
6. Make a test call (example below). If you get a rate-limit error, slow down or check the Limits page for the model you used.

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

The model name in the example is an alias. Check the model IDs your organization can use on the Limits page or in the models overview before you hard-code one.

Limits, spending caps and what happens when usage runs out

Mistral has three separate controls, and it helps to know which one stopped your requests.

Mistral also notes that if recent invoice payments fail, access can be affected until the payments go through. That does not apply to a Free mode account with no billing set up, but it matters once you add credits or a paid plan.

Pay-as-you-go: how to go beyond Free mode

Pay-as-you-go is not a separate plan. It is a billing setting that lets usage continue after your included allowance is used, and the extra usage draws from prepaid credits. You buy credits (or redeem a gift code) on the Billing page, then an admin enables pay-as-you-go. Because it is prepaid, your organization needs a credit balance before the setting can be switched on. Combined with the spending limit, this lets you cap your bill exactly.

What it costs after the free usage

Mistral prices the API per million tokens, with some exceptions: OCR is priced per 1,000 pages, speech models per minute and tool APIs per call. The pricing page also lists two discounts that matter for prototypes turning into products:

• Batch processing: prices are cut by 50% when you send work through the batch API instead of real-time calls.
• Cached input: repeated prompt content billed as cached input costs up to 90% less than normal input.
• Model choice: Mistral's own pricing FAQ recommends Mistral Medium for most tasks and coding, OCR for document AI, Voxtral for audio, and Mistral Small for cost-sensitive projects.

Per-model prices change with new releases, so check the live pricing page before you estimate costs. Paid Vibe plans (Pro, Education, Team) also include a monthly API credit allowance, which can be cheaper than buying credits if you already want the chat and coding tools. The pricing page lists Pro at $14.99 per month excluding taxes, and verified students can get it for $5.99 per month for up to 12 months.

Privacy: training on your API data

This is the part people most often miss. Mistral's help center has an article on opting out of training. For Mistral Studio and the API, it tells you to open Privacy in the left navigation of the Admin panel and turn off the Anonymous improvement data toggle. The article also says the Vibe (chat) setting and the API setting are separate, so turning off one does not turn off the other.

If you plan to send customer data, internal documents or anything confidential through a free account, check the Anonymous improvement data toggle first. We could not find an official statement of the current default for Free mode, so do not assume it is off.

Models you can try

The models overview lists Mistral's current lineup. Mistral does not publish a separate list of models that are free in Free mode; your Limits page shows what your organization can call. As of October 2026 the catalog includes:

Devstral and Magistral now appear only in Mistral's deprecated-models table, so avoid building new projects on them.

How Mistral's free tier compares

Mistral is one of the few big model makers that still gives API access with no card at all. Here is how it fits next to the other free API options we track:

If you want a single key that reaches many providers, OpenRouter's free variants are worth a look. If you mainly want strong general models with clear daily quotas, the Gemini API free tier is easier to plan around because Google lists which models are free and shows your per-model limits in AI Studio. For a full overview, see our guide to free AI API keys and how to get free AI credits.

Getting the most out of Free mode

• Pick the smallest model that works. Ministral and Mistral Small cost less per token once you pay, and on a rate-limited free plan they also respond faster, so you get more done per minute.
• Batch your evaluation runs. If you are testing prompts against a dataset, spread the calls out to stay under your requests-per-second limit instead of retrying failed calls in a tight loop.
• Use separate keys per project. Keys can have their own names and expiry dates, which makes it easy to revoke one leaked key without breaking everything else.
• Set a spending limit before adding credits. When you move to pay-as-you-go, set the organization monthly spending limit first so a bug cannot drain your balance.
• Watch the Limits page after model launches. Mistral adds and retires models often, and your available models and limits can change with them.

History and recent changes

• Earlier: Mistral's developer platform, then called La Plateforme, offered a free Experiment plan and a paid Scale plan. Help-center articles from that time mentioned phone verification and training on Experiment-plan data.
• 2026: the docs moved to the current structure: Free mode for new accounts, paid plans (Pro, Education, Team, Enterprise) with included monthly usage shared across Studio, the API and Vibe Code, and pay-as-you-go as a prepaid add-on. The console is now called Mistral Studio.
• Now (October 2026): Free mode is still active with no credit card required, and limits are still shown per organization rather than published.

Common mistakes

• Losing the key. Mistral shows the full key only once. If you close the dialog without copying it, create a new key.
• Trusting third-party numbers. Blog posts that quote exact free-tier limits often describe the old Experiment plan. Your Limits page is the only reliable source.
• Expecting overage to just work. Adding a card is not enough. Pay-as-you-go needs prepaid credits and must be switched on by an admin.
• Forgetting the second privacy toggle. The Vibe chat setting and the API setting are separate. Turning off training in one does not change the other.