Hugging Face's logo

qualcomm
/
Ministral-3-3B-Instruct-2512

Text Generation
PyTorch
llm
vlm
generative_ai
android
Model card Files
xet
 Community

Ministral-3-3B-Instruct-2512: Optimized for Qualcomm Devices

Ministral is a compact language model from Mistral AI designed for on-device deployment, offering strong instruction-following capabilities with low latency.

This is based on the implementation of Ministral-3-3B-Instruct-2512 found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.

Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up to run these models on a hosted Qualcomm® device.

Deploying Ministral-3-3B-Instruct-2512 on-device

Follow the GenieX quickstart to install GenieX and deploy the model on a target device.

Getting Started

There are two ways to deploy this model on your device:

Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

Runtime Precision Chipset SDK Versions Download
GENIEX_LLAMACPP q4_0 Universal Download

For more device-specific assets and performance metrics, visit Ministral-3-3B-Instruct-2512 on Qualcomm® AI Hub.

Option 2: Export with Custom Configurations

Use the Qualcomm® AI Hub Models Python library to compile and export the model with your own:

  • Custom weights (e.g., fine-tuned checkpoints)
  • Custom input shapes
  • Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for Ministral-3-3B-Instruct-2512 on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.text_generation

Model Stats:

  • Model architecture: Dense Transformer with Sliding Window Attention and Grouped Query Attention (GQA).
  • Response Rate: Rate of response generation after the first response token.
  • Supported languages: English, French, German, Spanish, Italian, Portuguese
  • TTFT: Time To First Token is the time it takes to generate the first response token. This is expressed as a range because it varies based on the length of the prompt.

Performance Summary

Model Runtime Precision Chipset Context Length Response Rate (tokens per second) Time To First Token (range, seconds)
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 512 24.451746 0.7207959999999999 - 2.8831839999999995
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 512 25.165428 0.397398 - 1.589592
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 512 23.007625 0.09381899999999999 - 0.37527599999999994
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 4096 8.491718 2.133671 - 68.277472
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 4096 16.567051 0.8275717741935483 - 26.482296774193546
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 4096 12.982683 0.11443629032258065 - 3.661961290322581
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 512 25.39869 0.8803166666666666 - 3.5212666666666665
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 512 24.879427 0.4110619999999999 - 1.6442479999999997
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 512 20.268594 0.10249566666666667 - 0.40998266666666666
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 4096 7.157775 2.314336935483871 - 74.05878193548388
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 4096 15.332285 1.035802129032258 - 33.14566812903226
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 4096 10.898818 0.15201735483870965 - 4.864555354838709
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 43.463984 0.21470266666666663 - 0.8588106666666665
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 24.93222 0.20670399999999997 - 0.8268159999999999
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 24.183669 0.08031866666666666 - 0.32127466666666665
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 27.059915 0.3847427096774193 - 12.311766709677418
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 20.453718 0.40758412903225805 - 13.042692129032257
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 16.015012 0.103447 - 3.310304
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 512 38.416646 0.3674443333333334 - 1.4697773333333335
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 512 29.349115 0.4189433333333333 - 1.6757733333333331
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 512 15.64538 0.18342466666666668 - 0.7336986666666667
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 4096 19.1478 0.8705155161290322 - 27.85649651612903
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 4096 17.635508 0.9817432903225806 - 31.415785290322578
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 4096 8.697967 0.25408487096774196 - 8.130715870967743
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 6.407106 3.153277 - 12.613108
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 3.117344 4.4006479999999994 - 17.602591999999998
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 2.426816 0.37938466666666665 - 1.5175386666666666
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 1.501261 5.474769774193549 - 175.19263277419356
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 2.448 6.569485129032258 - 210.22352412903226
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 2.081911 0.36488806451612904 - 11.67641806451613
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 512 38.416646 0.3674443333333334 - 1.4697773333333335
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 512 29.349115 0.4189433333333333 - 1.6757733333333331
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 512 15.64538 0.18342466666666668 - 0.7336986666666667
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 25.39869 0.8803166666666666 - 3.5212666666666665
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 24.879427 0.4110619999999999 - 1.6442479999999997
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 20.268594 0.10249566666666667 - 0.40998266666666666
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 4096 19.1478 0.8705155161290322 - 27.85649651612903
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 4096 17.635508 0.9817432903225806 - 31.415785290322578
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 4096 8.697967 0.25408487096774196 - 8.130715870967743
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 7.157775 2.314336935483871 - 74.05878193548388
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 15.332285 1.035802129032258 - 33.14566812903226
Ministral-3-3B-Instruct-2512 GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 10.898818 0.15201735483870965 - 4.864555354838709

License

  • The license for the original implementation of Ministral-3-3B-Instruct-2512 can be found here.

References

Community

Usage and Limitations

This model may not be used for or in connection with any of the following applications:

  • Accessing essential private and public services and benefits;
  • Administration of justice and democratic processes;
  • Assessing or recognizing the emotional state of a person;
  • Biometric and biometrics-based systems, including categorization of persons based on sensitive characteristics;
  • Education and vocational training;
  • Employment and workers management;
  • Exploitation of the vulnerabilities of persons resulting in harmful behavior;
  • General purpose social scoring;
  • Law enforcement;
  • Management and operation of critical infrastructure;
  • Migration, asylum and border control management;
  • Predictive policing;
  • Real-time remote biometric identification in public spaces;
  • Recommender systems of social media platforms;
  • Scraping of facial images (from the internet or otherwise); and/or
  • Subliminal manipulation
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
Text Generation
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Paper for qualcomm/Ministral-3-3B-Instruct-2512