Back to Models

Ministral 3B

mistralmistral/ministral-3b

API guide

Chat completion

Standard chat completions with system prompt and multi-turn conversation support. Stream responses for real-time UX.

cURL
curl https://api.therouter.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -d '{
    "model": "mistral/ministral-3b",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Classify this sentence: The cat sat on the mat."}
    ],
    "temperature": 0.1,
    "max_tokens": 256,
    "stream": true
  }'

Vision (image understanding)

Ministral 3 3B includes a 0.4B vision encoder supporting up to 10 images per prompt. Send images as base64 data URLs or remote URLs in the content array.

cURL
curl https://api.therouter.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -d '{
    "model": "mistral/ministral-3b",
    "messages": [
      {
        "role": "user",
        "content": [
          {"type": "text", "text": "Describe this image in detail."},
          {"type": "image_url", "image_url": {"url": "https://therouter.ai/assets/vision-sample.png"}}
        ]
      }
    ],
    "max_tokens": 1024
  }'

Function calling (tools)

Ministral 3 3B supports native function calling with tool definitions. Call functions, parse the tool call ID, execute, and return results β€” all within the standard chat completions endpoint.

cURL
curl https://api.therouter.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -d '{
    "model": "mistral/ministral-3b",
    "messages": [
      {"role": "user", "content": "What is the weather like in Paris?"}
    ],
    "tools": [{
      "type": "function",
      "function": {
        "name": "get_weather",
        "description": "Get the current weather for a city",
        "parameters": {
          "type": "object",
          "properties": {
            "city": {"type": "string", "description": "City name"}
          },
          "required": ["city"]
        }
      }
    }],
    "tool_choice": "auto",
    "temperature": 0.1
  }'

Structured JSON output

Use response_format to enforce structured JSON output, ideal for data extraction and classification pipelines.

cURL
curl https://api.therouter.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -d '{
    "model": "mistral/ministral-3b",
    "messages": [
      {"role": "user", "content": "Extract the name, date, and amount from this invoice."}
    ],
    "response_format": {
      "type": "json_schema",
      "json_schema": {
        "name": "invoice_extract",
        "strict": true,
        "schema": {
          "type": "object",
          "properties": {
            "name": {"type": "string"},
            "date": {"type": "string"},
            "amount": {"type": "number"}
          },
          "required": ["name", "date", "amount"],
          "additionalProperties": false
        }
      }
    },
    "temperature": 0.1
  }'
Fact ledger β€” every claim on this page traces here
sourceURLretrieved
Release dateAWS Bedrock model card β†—2026-05-27verified
ArchitectureHugging Face model card β†—2026-05-27verified
Context windowHugging Face model card β†—2026-05-27verified
Max output tokensAWS Bedrock model card β†—2026-05-27verified
Training cutoffβ€”β€”unknown
LicenseHugging Face β€” License β†—2026-05-27verified
LanguagesHugging Face β€” Languages β†—2026-05-27verified
Recommended temperatureHugging Face β€” Recommended settings β†—2026-05-27verified
VRAM requirement (FP8)Hugging Face β€” Deployability β†—2026-05-27verified
AIME 2025Hugging Face β€” Reasoning benchmarks β†—2026-05-27verified
AIME 2024Hugging Face β€” Reasoning benchmarks β†—2026-05-27verified
GPQA DiamondHugging Face β€” Reasoning benchmarks β†—2026-05-27verified
LiveCodeBenchHugging Face β€” Reasoning benchmarks β†—2026-05-27verified
Arena HardHugging Face β€” Instruct benchmarks β†—2026-05-27verified
WildBenchHugging Face β€” Instruct benchmarks β†—2026-05-27verified
MATH Maj@1Hugging Face β€” Instruct benchmarks β†—2026-05-27verified
MM MTBenchHugging Face β€” Instruct benchmarks β†—2026-05-27verified
Multilingual MMLUHugging Face β€” Base benchmarks β†—2026-05-27verified
MMLU (5-shot)Hugging Face β€” Base benchmarks β†—2026-05-27verified
Mistral 3 family announced β€” from 3B to 675BMistral AI Blog β†—2026-05-27verified
What is the difference between the Instruct and Reasoning variants?Hugging Face β€” Model variants β†—2026-05-27to verify
Can Ministral 3 3B run on a laptop?Hugging Face β€” Deployability β†—2026-05-27to verify
Does Ministral 3 3B support vision input?Hugging Face β€” Key Features β†—2026-05-27to verify
What license is Ministral 3 3B released under?Hugging Face β€” License β†—2026-05-27to verify
Help & contact