How to get a Mistral API key, and what the Mistral API costs
Updated · 5 min read
To get a Mistral API key, sign in to Mistral Studio at console.mistral.ai, add prepaid credits and turn on pay-as-you-go, then create a key under API Keys. There is no free API access any more: on 2026-11-18 Mistral's docs say the Free plan "doesn't include API access". Once billing is on, prices run from $0.10 per million tokens for the smallest model to $0.70 in and $2.10 out for the new Mistral Large 4. The API accepts OpenAI-style requests, so apps without a Mistral option can often use the key through a custom OpenAI-compatible provider.
Step by step
- Sign in at console.mistral.ai. Studio is Mistral's developer console (older guides call it La Plateforme). The same account is used for Vibe, Mistral's chat app formerly called Le Chat.
- Add credits. Open the Billing page in the Admin Panel (admin.mistral.ai/organization/billing), add a payment method and select Add credits. Credits are prepaid, and you need them before you can enable pay-as-you-go.
- Enable pay-as-you-go under Admin Panel → Subscription (admin.mistral.ai/subscription).
- Create the key at console.mistral.ai/api-keys: select New key, give it a name and an expiration date, and keep the connector scope at "Shared connectors only" unless the app needs your private connectors.
- Copy the key right away. Mistral shows the full key only once; if you lose it, delete it and make a new one.
- Paste it into your app and pick a model, for example
mistral-small-latest.
Is there a free tier?
No, not for the API. Free mode includes a monthly usage budget for trying Vibe and Studio in the browser, but the docs say "Creating API keys and using the API require billing activation and pay-as-you-go." This is a recent change. Mistral's help center still described Free mode as having the lowest limits "intended for evaluation and prototyping" in August 2026, and a pricing tracker recorded that the Free plan's $10 a month in API credits disappeared from Mistral's pricing page in mid-October 2026.
The closest thing to free usage is a paid plan. Paid plans include monthly usage shared across Studio, the API and Vibe Code, and that allowance is used before your credits. Mistral's pricing page lists Pro at $14.99 a month (before tax) with $25.5 a month in API credits. One model is free per token: Mistral Moderation 2. If you need a key that costs nothing at all, see free AI API keys.
Mistral API pricing
These were the standard prices per million tokens in Mistral's docs on 2026-11-18 (USD; the docs also show EUR):
| Model | API name | Input | Cached input | Output |
|---|---|---|---|---|
| Mistral Large 4 | mistral-large-latest |
$0.70 | $0.07 | $2.10 |
| Mistral Small 4 | mistral-small-latest |
$0.15 | $0.015 | $0.60 |
| Codestral | codestral-latest |
$0.30 | $0.03 | $0.90 |
| Ministral 3 14B | ministral-14b-latest |
$0.20 | $0.02 | $0.20 |
| Ministral 3 8B | ministral-8b-latest |
$0.15 | $0.015 | $0.15 |
| Ministral 3 3B | ministral-3b-latest |
$0.10 | $0.01 | $0.10 |
| DeepSeek V4.1 Flash (hosted by Mistral) | deepseek-v4-1-flash-1017 |
$0.30 | $0.007 | $1.20 |
| Z.ai GLM 5.3 (hosted by Mistral) | zai-glm-5-3 |
$1.40 | $0.14 | $4.40 |
Mistral Large 4 came out on 2026-11-05. It has a 1M-token context window, takes text and images, and Mistral publishes its weights. Mistral Small 4 has a 256k context window and Apache 2.0 open weights.
A chat turn of about 2,000 tokens in and 500 out costs roughly $0.0025 on Large 4 and $0.0006 on Small 4, so $10 of credit covers about 4,000 Large 4 turns. Batch jobs are billed at a 50% discount.
Mistral's main pricing page still quotes "Mistral Large" at $0.50 in and $1.50 out in its FAQ. That is the price of the older Mistral Large 3 (mistral-large-2512), which is deprecated. Mistral Medium 3.5, which the same page recommends, was also deprecated on 2026-10-31 and retires on 2027-04-30; Mistral names Large 4 as its replacement. OCR, transcription and speech models are priced per page, minute or character on Mistral's API pricing page.
Use the key in an app
If the app lists Mistral as a provider, choose it and paste the key. If it only has a custom OpenAI-compatible option, use:
- Base URL:
https://api.mistral.ai/v1 - API key: your Mistral key
- Model:
mistral-large-latest,mistral-small-latestor another name from the table
Mistral's migration guide says its Chat Completions API "follows the same request structure as OpenAI", and its tool calling uses the same schema. For code completion in an editor, Codestral answers fill-in-the-middle requests at https://api.mistral.ai/v1/fim/completions with the same key. Older tutorials describe a separate Codestral key and a codestral.mistral.ai endpoint; the current docs list only Studio, Vibe and legacy Mistral Code key types.
If you already have an OpenRouter key, OpenRouter lists Mistral Large 4 and Small 4 at the same per-token prices as Mistral.
Limits and gotchas
- Rate limits are set per model, as tokens per minute and requests per second. Mistral's docs don't list the numbers; yours are under Admin Panel → API → Limits. They rise automatically with your total billed amount, and support can raise them on request.
- When your credit balance reaches zero, API requests fail with HTTP 402. Credits are deducted including tax while spending limits apply before tax, so credits can run out before you reach your limit.
- A key belongs to one Workspace and uses that Workspace's quota and rate limits. For separate projects, create separate Workspaces and a key in each.
- Mistral retires model versions on a schedule. A dated name such as
mistral-medium-2508stopped working when that model retired on 2026-08-31; the models page lists deprecation and retirement dates.
Privacy and training
Mistral's help center says that on Free mode it may use your inputs and outputs to train its models, and that pay-as-you-go customers "retain full control over this processing and have the right to opt out at any time". It does not say whether pay-as-you-go accounts start opted in or out. To opt out, open the Admin Panel at admin.mistral.ai, go to Privacy and turn off the toggle under "Anonymous improvement data". The toggle for Vibe is separate, so set both if you use the chat app too.
Keep the key safe
- Give every key an expiration date and rotate keys regularly, as Mistral recommends.
- Use one key per app, so you can delete a single key if an app misbehaves.
- Never put the key in a website's front-end code or a public repository.
- Set a monthly spending limit on the Billing page before you give the key to a new app; when the limit is reached, API access can be suspended until the next month.
Apps that work with your Mistral API key




VisitSillyTavern
Free, open source · Self-hosted · Open source
Open-source LLM frontend for power users, popular for role-play and creative writing. Connects to many APIs with your own keys.
Works with:OpenAIAnthropicGoogle GeminiOpenRouter+3 more

VisitAnythingLLM
Free · macOS, Windows, Linux · Open source
Private AI app for chatting with your documents. Runs local models or connects to OpenAI, Anthropic, Gemini and others with your key.
Works with:OpenAIAzure OpenAIAnthropicGoogle Gemini+9 more
Other options
For more model choice behind one key, see the OpenRouter API key guide. For low per-token prices from the model's maker, see the DeepSeek API key guide. Prices across providers are compared in LLM API pricing, and Mistral's open-weight models can also run on your own hardware with the tools in local LLM apps.
Questions
Is there a free Mistral API key?
Not on the Free plan. On 2026-11-18 Mistral's docs say Free mode does not include API access, and creating API keys requires billing activation and pay-as-you-go. The Pro plan ($14.99 a month) includes $25.5 a month in API credits, which API calls use first.
How much does the Mistral API cost?
Per million tokens on 2026-11-18: Mistral Small 4 costs $0.15 input and $0.60 output, Mistral Large 4 $0.70 and $2.10, Codestral $0.30 and $0.90. Batch jobs are 50% off, and cached input on Mistral's own models costs a tenth of the normal input price.
Is the Mistral API compatible with OpenAI?
For chat, yes. Mistral's Chat Completions API follows the same request structure as OpenAI's, including tool calling, so an OpenAI client works with base URL https://api.mistral.ai/v1, your Mistral key and a Mistral model name.
Do I need a separate Codestral API key?
Mistral's current docs describe no separate Codestral key. Codestral (codestral-latest) runs on api.mistral.ai with your normal Studio key, including fill-in-the-middle completion at /v1/fim/completions.
Does Mistral train on my API data?
On Free mode, Mistral says it may use your inputs and outputs for training. Pay-as-you-go customers have the right to opt out at any time: in admin.mistral.ai, open Privacy and turn off the Anonymous improvement data toggle.
Sources
Checked against the providers' own documentation on 2026-11-18. Prices and limits change; the links below have the current figures.
- docs.mistral.ai/getting-started/quickstarts/studio/activate-and-generate-api-key
- docs.mistral.ai/admin/billing-usage/usage-limits
- docs.mistral.ai/admin/billing-usage/subscriptions
- docs.mistral.ai/admin/billing-usage/billing
- docs.mistral.ai/admin/identity-access/api-keys
- docs.mistral.ai/inference/pricing
- docs.mistral.ai/models
- docs.mistral.ai/models/mistral-large-4-0
- docs.mistral.ai/models/mistral-small-4-0-26-03
- docs.mistral.ai/models/codestral-25-08
- docs.mistral.ai/models/mistral-large-3-25-12
- docs.mistral.ai/inference/batch-processing
- docs.mistral.ai/resources/migration-guides
- mistral.ai/pricing
- help.mistral.ai/en/articles/698531
- help.mistral.ai/en/articles/347617-do-you-use-my-user-data-to-train-your-artificial-intelligence-models
- help.mistral.ai/en/articles/455207-can-i-opt-out-of-my-input-or-output-data-being-used-for-training
- www.usagepricing.com/blueprint/activity/mistral-ai-2026-10-15-vibe-api-credits-cut
- openrouter.ai/api/v1/models