Skip to content

Type a term, e.g. API, Shopware or SEO.

    ← All open AI models
    Free for commercial useRuns on: Laptop with 16 GBMistral AI · France

    Ministral 3 14B

    Ministral 3 14B is a small Mistral AI model with image input under the Apache 2.0 licence. Quantised it needs under 10 GB and runs on laptops and small computers, even without internet access. There are no independent scores from LMArena or Epoch yet; EuroEval places it mid-field in its German comparison.

    Profile

    Maker
    Mistral AI · France
    Licence
    Free for commercial use
    Runs on
    Laptop with 16 GB
    Memory at 4 bits
    9 GB
    Memory at 8 bits
    14 GB
    Parameters
    13.9bn
    Context
    100,000 tokens
    Inputs
    text, image
    Released
    December 2025

    Scores

    EuroEval
    German: rank 37 of 140

    Scores: LMArena, Epoch AI, EuroEval, Hugging Face, models.dev; as of 19 October 2026. Sources and method →

    Use cases

    Where the model makes sense.

    Assistant on every work laptop

    Situation
    Staff want to summarise emails and write drafts, but a central server is not worth it yet.
    What it looks like
    The model runs locally on the laptop, for example with LM Studio. Everyone uses their own copy and nothing leaves the device.
    What to watch
    Keep versions consistent so everyone works with the same model and results are comparable.

    AI at the machine without internet

    Situation
    In production, computers at machines may not connect to the internet for security reasons.
    What it looks like
    The model runs on the machine PC and explains fault messages using the operating manual stored beforehand.
    What to watch
    For safety-relevant instructions the maker’s manual always applies, not the summary.

    Pre-sorting for a larger model

    Situation
    Many requests are simple; a few need a large model or a person.
    What it looks like
    Ministral pre-sorts: it answers simple questions itself and forwards difficult ones to a larger model or the team.
    What to watch
    Test the routing rules on real examples before going live.

    Reading labels and notices

    Situation
    In the warehouse, details from labels and notices should be captured quickly.
    What it looks like
    Take a photo with a tablet; the model returns item number, quantity and date as text.
    What to watch
    For barcodes and QR codes a classic scanner is more reliable than a language model.

    The use cases were written with AI assistance (Claude by Anthropic) and reviewed by vona. Scores come unchanged from the sources listed.

    When another model fits better

    For demanding German texts and analysis, Mistral Small 3.2 is a better fit once a computer with 32 GB is available.

    How to start

    ollama run ministral-3:14b

    With Ollama or LM Studio. On laptops with 16 GB the 4-bit version is right; the smaller Ministral 3 8B is even more frugal.

    Suitable interfaces and programs from our directory: Ollama, LM Studio

    Licence: Free for commercial use

    The licence allows commercial use, modification and redistribution; the main obligation is to keep the licence notice.

    Hugging Face ↗

    Similar models

    Comparable models.

    • Mistral AI · France

      Ministral 3 8B

      An even smaller version of Ministral 3 with image input that gets by with 16 GB of memory.

      Free for commercial use · Runs on: Laptop with 16 GB

      In the overview →
    • Mistral AI · France

      Mistral Small 3.2

      Compact Mistral AI model under the Apache 2.0 licence that runs on a computer with 32 GB. It ranks near the top of EuroEval's German comparison.

      Free for commercial use · Runs on: Computer with 32 GB

      Use cases and details →
    • Google · United States

      Gemma 4 E4B

      Small Gemma model for laptops and high-end phones that accepts text, images and audio.

      Free for commercial use · Runs on: Laptop with 16 GB

      Use cases and details →

    FAQ

    What hardware do I need for Ministral 3 14B?

    In the common 4-bit version, Ministral 3 14B needs about 9 GB of memory including headroom for context and runtime. That fits: Laptop with 16 GB. The more accurate 8-bit version needs about 14 GB. Long documents increase the requirement.

    May I use Ministral 3 14B commercially?

    The licence allows commercial use, modification and redistribution; the main obligation is to keep the licence notice.

    How good is Ministral 3 14B in German?

    In EuroEval’s German comparison, Ministral 3 14B ranks 37 of 140. There are no human ratings in German on LMArena yet; test it with your own examples.

    AI consulting

    Your own language model for your company?

    Together we find the model that fits your tasks, your hardware and your data protection requirements, test it with real documents and set up operation.

    Newsletter

    News from the vona workshop.

    AI tools we actually use, our weekly AI review & insights from our projects – short, practical and without spam.