Hugging Face's logo

qualcomm
/
LFM3-VL-2B

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

LFM3-VL-2B: Optimized for Qualcomm Devices

LFM3-VL is a vision-language model from Liquid AI designed for fast, high-quality on-device deployment. It understands text and images for tasks such as visual question answering, image captioning, and optical character recognition.

This is based on the implementation of LFM3-VL-2B 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 LFM3-VL-2B 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 LFM3-VL-2B 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 LFM3-VL-2B on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.text_generation

Model Stats:

Performance Summary

Model Runtime Precision Chipset Context Length Response Rate (tokens per second) Time To First Token (range, seconds)
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 512 50.719403 0.2964213333333333 - 1.1856853333333333
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 512 44.960191 0.17246033333333333 - 0.6898413333333333
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 512 50.390189 0.038746333333333334 - 0.15498533333333334
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 4096 40.378383 0.6042843870967741 - 19.337100387096772
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 4096 39.570671 0.20865670967741934 - 6.677014709677419
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 4096 44.708442 0.05225441935483871 - 1.6721414193548387
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 512 51.522386 0.3240826666666667 - 1.2963306666666667
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 512 45.54272 0.1871413333333333 - 0.7485653333333332
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 512 40.336723 0.048272666666666665 - 0.19309066666666666
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 4096 36.674873 0.5774609677419356 - 18.478750967741938
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 4096 36.384082 0.25551380645161287 - 8.176441806451612
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 4096 37.143955 0.053223806451612905 - 1.703161806451613
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 87.335232 0.09055366666666666 - 0.36221466666666663
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 59.68671 0.088995 - 0.35598
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 55.52507 0.03872633333333333 - 0.1549053333333333
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 70.601408 0.13769319354838708 - 4.4061821935483865
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 54.861289 0.1042028064516129 - 3.3344898064516126
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 44.581082 0.04543903225806451 - 1.4540490322580644
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 512 56.998199 0.31141599999999997 - 1.2456639999999999
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 512 44.746186 0.188113 - 0.752452
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 512 33.711526 0.089683 - 0.358732
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 4096 54.744065 0.3849749032258064 - 12.319196903225805
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 4096 40.563722 0.23184312903225804 - 7.418980129032257
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 4096 31.365092 0.09020838709677419 - 2.886668387096774
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 18.427702 1.4237846666666667 - 5.695138666666667
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 7.673146 2.118666333333333 - 8.474665333333332
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 27.986364 0.1588993333333333 - 0.6355973333333332
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 9.725378 1.7475810967741936 - 55.922595096774195
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 7.083885 2.377958322580645 - 76.09466632258064
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 24.978056 0.15768970967741935 - 5.046070709677419
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-9075 512 40.919212 0.5875876666666666 - 2.3503506666666665
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-9075 512 18.409251 0.7199803333333332 - 2.879921333333333
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-9075 512 41.93787 0.100407 - 0.401628
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 512 56.998199 0.31141599999999997 - 1.2456639999999999
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 512 44.746186 0.188113 - 0.752452
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 512 33.711526 0.089683 - 0.358732
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 51.522386 0.3240826666666667 - 1.2963306666666667
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 45.54272 0.1871413333333333 - 0.7485653333333332
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 40.336723 0.048272666666666665 - 0.19309066666666666
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-9075 4096 35.48255 0.6813493225806452 - 21.803178322580646
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-9075 4096 17.043152 0.8256118064516129 - 26.419577806451613
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-9075 4096 39.31402 0.10048738709677418 - 3.215596387096774
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 4096 54.744065 0.3849749032258064 - 12.319196903225805
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 4096 40.563722 0.23184312903225804 - 7.418980129032257
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 4096 31.365092 0.09020838709677419 - 2.886668387096774
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 36.674873 0.5774609677419356 - 18.478750967741938
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 36.384082 0.25551380645161287 - 8.176441806451612
LFM3-VL-2B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 37.143955 0.053223806451612905 - 1.703161806451613

License

  • The license for the original implementation of LFM3-VL-2B 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