Mistral Large

Mistral Large is a text generation model from Mistral, positioned by the provider as its flagship model. The version served here is Mistral Large 2 (mistral-large-2407), a proprietary weights-available model that Mistral describes as strong at reasoning, code, JSON and chat. It was released by the provider on 26 February 2024, and its knowledge cutoff is 30 November 2024, so anything that happened after that date is outside what the model learned during training.

The model accepts text and file inputs and returns text. It has a context window of 128,000 tokens, which is enough to hold long documents, large code files or extended conversations in a single request, and it can produce up to 102,400 tokens in one response. Because files are an accepted input modality, you can pass document content alongside your instructions rather than pasting everything into the prompt yourself. Output is text only, so if you need images or audio you will need a different model.

Those capabilities point to a fairly broad set of practical uses. The long context and reasoning strength suit document analysis, summarisation and question answering over lengthy source material, including contracts, reports and research papers supplied as files. The provider highlights code and JSON, so code explanation, generation and refactoring tasks are reasonable fits, as is producing structured JSON payloads for downstream systems. Chat and assistant-style applications also work well, particularly where the conversation needs to carry a lot of prior context without truncation.

You call the model through api14's OpenAI-compatible API using the model id "mistralai/mistral-large". Because the endpoint follows the OpenAI schema, existing OpenAI client libraries and request shapes work with only the base URL and model id changed. The supported request parameters are frequency_penalty, max_tokens, presence_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools and top_p. Tool calling is available through tools and tool_choice, and structured_outputs plus response_format let you constrain the shape of the returned text when you need reliable JSON. The seed parameter helps when you want repeatable sampling, and stop, temperature and top_p give you control over where and how the model generates.

Keep the knowledge cutoff in mind: for questions about events after 30 November 2024, the model has no built-in knowledge and will need that information supplied in the prompt or through tools. The 128,000 token context and 102,400 token output ceiling are generous but finite, so very large corpora still need chunking or retrieval. If your task is not text based, or if you specifically need a model with a more recent cutoff, another model on api14 may be the better choice.