Mistral: Mistral Small 4 (batch)

Mistral Small 4 (batch) is a text generation model from Mistral, listed on api14 under the model id mistralai/mistral-small-2603:batch. It is the next major release in the Mistral Small family, and according to Mistral it unifies the capabilities of several flagship Mistral models into a single system, combining strong reasoning with the other strengths of those models. The batch label marks it as the batch variant of this release. Mistral is the provider behind the model, and the provider release date is 16 March 2026.

On the input side, the model accepts text and images, and it produces text as output. Its context window is 262,144 tokens, which allows very long documents, transcripts or multi file prompts to be handled in a single request. The maximum output length is 209,715 tokens, so long form generations are possible alongside long inputs. The parameter list includes a reasoning flag and a reasoning effort control, plus include_reasoning, which indicates the model can expose its reasoning process when asked. It also supports structured outputs and response_format, so replies can be constrained to a chosen shape rather than free text.

Those capabilities point to a set of practical uses. Long context plus image input suits document analysis where a developer sends scanned pages, screenshots or diagrams together with a written question and expects a text answer. The reasoning controls fit tasks such as multi step analysis, planning or problem solving where extra deliberation improves the result. Tool and function support (tools, tool_choice) makes it usable as the decision layer in an agent that calls external services. Structured outputs and response_format are useful when the model must emit JSON or another machine readable structure for downstream code, and the generous output ceiling helps with summarisation, report drafting or code generation over large inputs.

You call the model through api14's OpenAI-compatible API using the model id mistralai/mistral-small-2603:batch in the usual chat completions style request. The supported 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. Seed gives reproducible sampling when you need consistent outputs across runs, while stop sequences, temperature and top_p let you control where and how generation ends. Because the interface is OpenAI-compatible, existing client libraries and SDKs generally work after you swap in the api14 base URL and this model id.

One limit worth noting is modality: input can include images, but output is text only, so the model is not a fit for image generation. If your workload is purely text, a non batch sibling in the Mistral Small family may be worth comparing on the live price table, but the batch variant here remains a reasonable choice when you want the combined reasoning capabilities with a very large context window and structured outputs.