Mistral as a drop-in for OpenAI

Mistral's /v1/chat/completions endpoint accepts the same JSON schema as OpenAI's, so most SDKs and frameworks work by swapping the base URL and API key. Here's the mapping.

1. Base URL & auth

OPENAI_API_BASE=https://api.mistral.ai/v1
OPENAI_API_KEY=<your-mistral-key>   # get one at console.mistral.ai
Authorization: Bearer <your-mistral-key>

2. OpenAI Python SDK

from openai import OpenAI

client = OpenAI(
    base_url="https://api.mistral.ai/v1",
    api_key="<your-mistral-key>",
)

resp = client.chat.completions.create(
    model="mistral-large-latest",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(resp.choices[0].message.content)

3. curl

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

4. Model-name mapping

5. Supported endpoints

• /v1/chat/completions — full parity, including streaming and tool calls.
• /v1/embeddings — model mistral-embed.
• /v1/models — list available models on your key.
• Not mirrored: Assistants, vision inputs on OpenAI schema, audio, image gen. Use Mistral's native SDK for agents and file inputs.

6. Common gotchas

• Response format: use { "type": "json_object" } — same as OpenAI.
• System prompts: supported, but Mistral models weight them slightly differently — retest your prompts.
• Tool choice: tool_choice: "auto" | "any" | "none" — "any" replaces OpenAI's "required".
• Rate limits: per-key and per-model, in requests/min and tokens/min — see console.mistral.ai → Usage.

FAQ