Mistral: Ministral 3 3B 2512

Mistral: Ministral 3 3B 2512 is a text generation model from Mistral, released on 2 December 2025. It is the smallest member of the Ministral 3 family, which Mistral describes as a line of efficient tiny language models. The model id on api14 is mistralai/ministral-3b-2512, and its main category is text.

The model accepts text and image input and produces text output, so it can handle both plain language prompts and prompts that include images. It has a context window of 131,072 tokens and a maximum output of 104,857 tokens, which means you can send long documents or long conversations and still leave room for a substantial generated response. Supported request parameters include temperature, top_p, frequency_penalty, presence_penalty, max_tokens, stop, seed, response_format, structured_outputs, tools and tool_choice. The presence of tools and tool_choice means it can take part in function calling or tool use flows, and structured_outputs plus response_format let you ask for machine readable responses rather than free text.

Because it handles both text and images, practical uses include describing or extracting information from images, answering questions about screenshots or diagrams, and mixed prompts where an image is accompanied by written instructions. Its long context makes it suitable for summarising long documents, working through lengthy chat histories, or processing large batches of text in a single request. Tool support makes it a candidate for agent style workflows where the model decides when to call an external function, and structured outputs help when the result needs to feed into another program, such as filling a JSON object or a fixed schema. The large maximum output also suits tasks that require long generated answers, such as drafting reports or expanding notes into full text.

You call the model through api14's OpenAI-compatible API using the model id mistralai/ministral-3b-2512 in the same way you would call any other chat completions style endpoint. Set the model field to that id, provide your messages, and add any of the supported parameters you need: temperature, top_p, frequency_penalty, presence_penalty, max_tokens, stop, seed, response_format, structured_outputs, tools and tool_choice. Because the interface is OpenAI-compatible, existing OpenAI client libraries and request shapes generally work without changes, which keeps migration effort low.

Since this is the smallest model in its family, it is a reasonable choice when latency or throughput matters more than maximum reasoning depth. For tasks that need heavier reasoning or more complex multi step planning, a larger model in the same family or another provider may fit better. The facts here do not state benchmark results, so evaluate it against your own workload before committing.