Intermediate
La Plateforme
La Plateforme is Mistral's developer API — the technical entry point for building applications with Mistral models. It provides pay-per-token access to all Mistral models via an OpenAI-compatible REST API.
API Pricing (May 2026)
| Model | Input / 1M tokens | Output / 1M tokens | Context | Best for |
|---|---|---|---|---|
| mistral-large-latest (Large 3) | $2.00 | $6.00 | 128K | Complex reasoning, long documents |
| mistral-medium-latest (Medium 3.5) | $1.50 | $7.50 | 128K | Frontier quality, agentic tasks |
| mistral-small-latest (Small 4) | $0.15 | $0.60 | 128K | High-volume, cost-efficient tasks |
| codestral-latest | $0.30 | $0.90 | 256K | Code generation — very long context |
| devstral-small | $0.07 | $0.28 | 128K | Agentic coding — cheapest agentic model |
| pixtral-large | $2.00 | $6.00 | 128K | Vision — image + text understanding |
| mistral-nemo | $0.02 | $0.06 (approx) | 128K | Ultra-cheap tasks — lowest cost per token |
| mistral-embed | $0.10 | — | 8K | Text embeddings for RAG |
Quick Start
pip install mistralai
from mistralai import Mistral
client = Mistral(api_key="your_api_key")
response = client.chat.complete(
model="mistral-large-latest",
messages=[{"role": "user", "content": "Explain RAG in 3 bullet points"}]
)
print(response.choices[0].message.content)OpenAI-Compatible Endpoint
La Plateforme exposes an OpenAI-compatible API at https://api.mistral.ai/v1/. Any code using the OpenAI Python SDK can talk to Mistral with one change — the base URL and API key:
from openai import OpenAI
client = OpenAI(
api_key="your_mistral_api_key",
base_url="https://api.mistral.ai/v1/"
)
response = client.chat.completions.create(
model="mistral-large-latest",
messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)Key Platform Features
Function Calling / Tool Use
All main models support tool calling — define functions in JSON schema, the model decides when to call them. Used to build agents on La Plateforme.
JSON Mode
Force output to be valid JSON — useful for structured data extraction. Set response_format: {type: 'json_object'}.
Streaming
All models support streaming responses. Use stream=True to get tokens as they generate for lower perceived latency.
Batch API
Submit large batches of requests (up to 50,000 per batch) at 50% discount. Results returned within 24 hours.
Fine-tuning
Fine-tune Mistral Small or Mistral 7B on your own data. Upload a JSONL file, train, and deploy the fine-tuned model via API.
EU Data Residency
API calls can be routed to EU infrastructure (Paris). All data stays within the EU — important for GDPR compliance.
Free Tier
La Plateforme has a free tier with rate limits — useful for experimenting and prototyping without adding a payment method. Limits: typically 1 request/second and 500,000 tokens/month on free tier (check current limits at console.mistral.ai).
Checklist: Do You Understand This?
- Can you name the cheapest and most capable Mistral models with approximate prices?
- Can you call La Plateforme using the OpenAI Python SDK with a base URL change?
- Do you know how to enable streaming or JSON mode in a Mistral API call?
- Do you understand what the batch API offers and its discount rate?