Available now
Mistral
Mistral

ministral-3b-2512

The smallest model in the Ministral 3 family, Ministral 3 3B is a powerful, efficient tiny language model with vision capabilities.

TextToolsFilesOpen WeightsVision131.1K
InputFree
OutputFree
Context131.1K
Endpointsopenai

Capabilities

VisionStructured

Modalities

Input
textimage
Output
text

Quick stats

Context window131.1K
Max output131.1K
Modechat
TokenizerMistral
Quantizationfp8

Performance

Loading performance data...

Supported parameters

ParameterAlwaysDefault
frequency_penalty(do not send)
logprobs--
max_tokens-
presence_penalty-
repetition_penalty--
response_format-
seed-
stop-
structured_outputs-
temperature0.3
tool_choice-
tools-
top_logprobs--
top_p(do not send)
§ 01

Pricing

Input price$0.00 · 1M tokens
Output price$0.00 · 1M tokens
Context window131.1K tokens
Compatible endpointsopenai
VendorMistral
§ 02

Call ministral-3b-2512 from your code

Point any OpenAI-compatible SDK at UnoRouter and request the model by name. Replace YOUR_API_KEY with a real key from your dashboard.

bash
curl https://api.unorouter.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "ministral-3b-2512",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Sign in to auto-fill your API key

§ 03

Frequently asked questions

How much does ministral-3b-2512 cost per 1M tokens?

Input is priced at $0.00 per 1M tokens, output at $0.00 per 1M tokens. Billing is per token, no rounding to batch sizes.

How do I access ministral-3b-2512 via API?

Send requests to the UnoRouter /v1/chat/completions endpoint with model=ministral-3b-2512. Any OpenAI-compatible client library works. Authentication uses a standard Bearer token.

What is the context window of ministral-3b-2512?

ministral-3b-2512 supports a context window of 131.1K tokens, shared between your prompt and the model's response.

§ 04

Similar models

Try ministral-3b-2512 now

Create an API key and start making requests in under a minute.

View all models