Mistral Large 2512 Batch
Neural Network
Mistral Large 2512 is Mistral AI's flagship sparse MoE language model, available via API in BotHub.
Max answer length
(in tokens)
Context size
(in tokens)
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Mistral Large 2512 Batch?
Mistral Large 3 2512 is developed by the French lab Mistral AI and is their most capable model at the time of release. It is based on a sparse mixture-of-experts architecture: out of 675 billion parameters, 41 billion are activated per request, combining large model quality with efficient token processing. The model is released under the Apache 2.0 license, and this version is designed for batch processing where volume matters more than instant response. In BotHub, you connect from Russia without a VPN or foreign card, pay in rubles for tokens used rather than a subscription, and work with privacy: traffic is encrypted, and chats are not archived. A unified OpenAI-compatible API allows switching to any of the 250+ available models without rewriting integration, while companies get access to contracts, invoices, EDI, and an admin panel with limits. Useful for mass processing documents and requests, parsing logs, writing code, analyzing large text corpora, building function-calling scenarios, or running long reasoning chains on a schedule.Frequently asked questions about Mistral Large 2512 Batch
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
This is a Mistral text model with a 262,144 token context: it parses long documents, writes and refactors code, and helps with debugging, analytics, and STEM tasks. It accepts not only text but also images and documents as input, and can call external functions.
The single response limit is 209,715 tokens, which is a very large amount of text: enough for a long report, documentation, or an entire large code module. Keep in mind that the 262,144 token window is shared between the request and response, so leave a buffer for huge inputs.
Our data does not indicate a separate reasoning mode, but that doesn't mean the model can't handle multi-step logic. Just ask it to break down the solution step-by-step or formulate a plan first, and you will get a detailed analysis, not just the final conclusion.
Yes, function calling and structured JSON output are confirmed. The model will select the necessary tool, insert arguments, and return the response strictly according to your schema — this is convenient for agents and integrations. Connection is via the unified OpenAI-compatible BotHub API, so you can switch models without rewriting code.
According to our data, the model accepts images and documents, and works with documents natively: you can send a scan, diagram, or report and ask it to analyze the content. Audio and video are not listed in the input formats — BotHub has specialized models for those in the same window.
There are no notes about languages in our data, so we cannot give guarantees. The easiest way to check is through practice: send your typical request — a contract, correspondence, or ticket — and compare it with other models. Payment in BotHub is per token, so such a test does not require a subscription.