Mistral makes some of the best open-weight models available. Getting an API key takes about two minutes, and you don't need a credit card.
What you need before getting started
You need:
- An email address (or a Google, GitHub, Microsoft, or Apple account to sign up faster)
- A phone number that can receive SMS, if you sign up with email
No credit card required for Mistral's default Free mode.
One thing worth knowing upfront: on Free mode, Mistral may use your API inputs and outputs to train its models unless you opt out. Pay-as-you-go accounts get a training opt-out toggle and can request Zero Data Retention for supported stateless endpoints. See how Mistral retains your data for the full breakdown before you send anything sensitive.
How to generate your Mistral API key (step by step)
- Go to console.mistral.ai, Mistral's Studio console, and create an account
- Verify your phone number if prompted. This is how Mistral gates Free mode for email signups
- Once inside Studio, navigate to API Keys in the left sidebar
- Click Create new key
Source: Mistral Docs, Activate Studio and generate an API key, captured September 2026
- Give the key a name, set an expiration date (Mistral now prompts for one on every new key), choose a connector access scope, and confirm
Source: Mistral Docs, Activate Studio and generate an API key, captured September 2026
- Copy the key immediately and store it somewhere safe. Mistral won't show it again
Mistral API rate limits on the free plan
Mistral no longer publishes fixed free-tier numbers on its docs. Free mode enforces three limits per Organization: requests per second, tokens per minute, and tokens per month, and Mistral says these are intentionally conservative, meant for evaluation and prototyping rather than production traffic. Check your exact numbers anytime on the Limits page in the Studio Admin Panel.
In practice, tokens per minute is the one you'll hit first if you're batching requests or running anything concurrent. If a request gets rate limited, the API returns a 429 and you'll need to retry with backoff.
If you outgrow Free mode, switching on pay-as-you-go removes the evaluation-tier ceilings. Limits then scale automatically with usage tiers (Tier 1 through Tier 4+) based on your cumulative billed amount, not a flat upgrade to a named plan. Enabling pay-as-you-go unlocks Tier 1 immediately; crossing $20, $100, and $500 billed lifts you to Tiers 2 through 4.
Mistral API pricing: how much does it cost
Mistral charges per million tokens, counting both input (what you send) and output (what comes back). There's no subscription fee on the API side. You pay for what you use.
| Model | Input | Output |
| Mistral Small 4 | $0.15 / M tokens | $0.60 / M tokens |
| Mistral Large 3 | $0.50 / M tokens | $1.50 / M tokens |
To put those numbers in context: a million input tokens is roughly 750,000 words, or about 1,500 pages of text. At Mistral Small's rates, that costs fifteen cents. For most personal or small-team use, the monthly bill is negligible.
Mistral raised Small's per-token rate in July 2026 alongside the Small 3.2 to Small 4 upgrade, so budget a little higher than older guides suggest. Two ways to cut the bill further: batch API requests that don't need an immediate response get a 50% discount, and cached input tokens (repeated prompt prefixes) can cost up to 90% less than fresh input tokens.
Which Mistral model should you use
If you're not sure, use mistral-small-latest.
It's fast, cheap, multimodal, multilingual, and Apache 2.0 licensed. The Apache 2.0 license matters if you're building something commercial. No usage restrictions, no license fees.
Step up to mistral-large-latest if you need heavier reasoning: complex multi-step tasks, nuanced analysis, or anything where output quality is worth paying more for. It's now a little over three times Small's input price and 2.5x its output price, closer than it used to be, but still worth being deliberate about when you actually need it.
How to use your Mistral API key in your project
Set it as an environment variable so it never gets hardcoded into your source files:
export MISTRAL_API_KEY="your-key-here"
Test it with a curl call:
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": "Hello"}] }'
If you get a JSON response with a choices array back, the key is working. If you get a 401, double-check that the environment variable is set correctly in your current session.
For Python projects, the python-dotenv package is the standard way to load a .env file at runtime rather than setting variables manually each session:
from dotenv import load_dotenv
import os
load_dotenv()
api_key = os.getenv("MISTRAL_API_KEY")
Use your Mistral API key in Anarlog for AI meeting notes that stay on your device
Getting your own API key is a deliberate choice. It means you want control over what AI you're using and what happens to the data you send it. Anarlog works on the same principles.
It’s an open-source AI notepad for meetings that gives you complete control over your AI stack and your data.
The workflow is simple: record, transcribe locally, summarize with your own Mistral key. You choose which model runs. You can swap to Anthropic, OpenAI, or a local Ollama model any time without losing your files or your history.
The local meeting record, audio, transcript, and summary remain on your device by default. Anarlog stores meeting data in local SQLite; if you enable cloud AI, CloudSync, or sharing, assess those services independently for the data you send to them.
Plus, all core features, local transcription, and BYOK stay completely free. You’re already paying for the API key, you shouldn’t have to pay twice. But if you want cloud services and don't want to manage keys at all, there is a $15/month plan you can check out.
To connect Mistral, open Anarlog's settings, go to API Keys, paste your key, and that's it. Try it out for free now - Download Anarlog.