Ministral 3 14B 2512

Mistral AI's Ministral 3 14B is the largest entry in the Ministral 3 family and is positioned to deliver capabilities and performance comparable to the larger Mistral Small 3.2 24B, while being optimized for running on local hardware rather than only in the cloud. It is distributed under the Apache 2.0 license as a generally available release, and Mistral pairs the model page with a technical report on arXiv so practitioners can study the design choices behind the family. The model accepts both text and image inputs and produces text output, and is documented to support structured outputs and function calling through Mistral's chat and conversations endpoints, giving developers predictable ways to wire it into agentic and tool-using applications.

Because it is tuned for on-device and diverse-hardware deployment, Ministral 3 14B suits use cases where a balance of capability and portability matters more than chasing the largest possible parameter count, such as local assistants, private document and image workflows, and embedded retrieval or tool-calling pipelines. The Mistral 3 release framing positions this 14B variant as a strong middle ground that narrows the gap to bigger Mistral Small models while remaining light enough to run outside of a datacenter. Its general-availability Apache 2.0 status and accompanying technical report also make it attractive for teams that want both open licensing and a reference document describing the model's design intent.