Mistral: Mistral Small 4

Mistral: Mistral Small 4 is a text generation model from Mistral, released on 16 March 2026. It is the next major release in the Mistral Small family, and Mistral describes it as unifying the capabilities of several flagship Mistral models into a single system, combining strong reasoning with the other strengths of those models. On api14 it sits in the text category and is identified by the model id mistralai/mistral-small-2603.

The model accepts text and images as input and returns text. Its context window is 262,144 tokens, and the maximum output is 209,715 tokens, so both very long inputs and long generations are possible within the same request. Because image input is supported, prompts can mix written instructions with images, while the response itself is always text.

Those capabilities point to a fairly wide set of practical uses. Long context suits work such as summarising or querying large documents, codebases, transcripts or report sets in a single call, and the large output ceiling helps when the answer itself needs to be long, for example a full draft or an extended analysis. Image input makes it usable for tasks that combine a picture with written instructions, such as describing or extracting information from a screenshot or diagram. Reasoning support and tool calling make it a reasonable fit for agent style workflows and for applications that need to call functions or external services, and structured outputs help when the result has to be parsed by another program.

You call it through api14's OpenAI-compatible API using the model id mistralai/mistral-small-2603, so existing OpenAI client libraries and request shapes work with only the model name changed. The supported request parameters are frequency_penalty, include_reasoning, max_tokens, presence_penalty, reasoning, reasoning_effort, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools and top_p. In practice that covers sampling controls, deterministic runs via seed, stopping conditions, JSON or schema constrained responses through response_format and structured_outputs, and tool use through tools and tool_choice, with reasoning controlled by the reasoning, reasoning_effort and include_reasoning options.

There are no stated restrictions in the facts above beyond the modality split: input can include images, but output is text only, so it is not a fit if you need generated images, audio or video. If your workload is purely short text prompts with no images, no tool calls and no need for a long context, a smaller and simpler text model may be sufficient, but for long inputs, long outputs, image aware prompts or tool driven pipelines, this model covers those cases directly.