AI BriefingMistral-Large-4-1T-A52B: Mistral Large 4 with 1T parameters and 52B active parameters
mistralai/Mistral-Large-4-1T-A52B
About the project
Mistral Large 4 is a hybrid model with a total of 1T parameters, activating only 52B during inference. It is designed to hold a vast amount of knowledge while keeping actual computational costs close to those of smaller models.
The chat template allows direct control over tool calls, JSON schema-based response formats, and reasoning effort levels. Reasoning depth can be adjusted across three levels—low, medium, and high—to balance response speed and accuracy.
It supports multimodal processing by accepting text, images, and image URLs as input. It is well-suited for complex agent tasks by precisely placing tokens according to the conversation flow, including system prompts, user messages, and tool results.
It is suitable for developers who expect high performance on tasks requiring large-scale knowledge but want to avoid increased inference costs due to the number of active parameters. It can be used to build applications that require structured output and tool integration.
mistralai/Mistral-Large-4-1T-A52B
The original page has no description.
image-text-to-text
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.