Hugging Face's logo

qualcomm
/
LFM2.5-1.2B-Instruct

Text Generation
PyTorch
llm
generative_ai
android
Model card Files
xet
 Community

LFM2.5-1.2B-Instruct: Optimized for Qualcomm Devices

LFM2.5 is a family of hybrid language models from Liquid AI designed for fast, high-quality on-device deployment. The 1.2B Instruct model is tuned for instruction following, chat, and tool use.

This is based on the implementation of LFM2.5-1.2B-Instruct 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 LFM2.5-1.2B-Instruct 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 LFM2.5-1.2B-Instruct 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 LFM2.5-1.2B-Instruct on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.text_generation

Model Stats:

  • Model architecture: Hybrid architecture combining short-range gated convolutions with grouped-query attention (LFM2).
  • Response Rate: Rate of response generation after the first response token.
  • Supported languages: English, Arabic, Chinese, French, German, Japanese, Korean, Spanish

Performance Summary

Model Runtime Precision Chipset Context Length Response Rate (tokens per second) Time To First Token (range, seconds)
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 512 80.260596 0.20045933333333335 - 0.8018373333333334
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 512 68.165939 0.10517033333333334 - 0.42068133333333335
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 512 68.33336 0.026763333333333337 - 0.10705333333333335
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 4096 59.144475 0.44533467741935484 - 14.250709677419355
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 4096 60.944394 0.14961758064516129 - 4.787762580645161
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Gen 5 For Galaxy Mobile 4096 64.29359 0.03694151612903225 - 1.182128516129032
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 512 80.363343 0.22575 - 0.903
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 512 72.259121 0.11525599999999998 - 0.46102399999999993
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 512 56.965983 0.03423233333333333 - 0.13692933333333332
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 4096 45.639921 0.48976935483870965 - 15.672619354838709
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 4096 57.849112 0.16003048387096772 - 5.120975483870967
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite For Galaxy Mobile 4096 46.243552 0.0413068064516129 - 1.3218178064516128
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 137.064045 0.20856166666666665 - 0.8342466666666666
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 90.203988 0.06708733333333333 - 0.26834933333333333
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 81.831282 0.026624 - 0.106496
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 109.706018 0.11465490322580645 - 3.6689569032258063
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 68.126927 0.09095838709677419 - 2.910668387096774
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 67.862064 0.03592683870967742 - 1.1496588387096773
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 28.083614 0.9435203333333334 - 3.7740813333333336
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 11.860398 1.3878993333333334 - 5.5515973333333335
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 41.651454 0.08812133333333333 - 0.3524853333333333
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 15.47093 1.2430317419354837 - 39.77701574193548
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 9.389539 1.6565857096774192 - 53.010742709677416
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 30.265886 0.09416838709677418 - 3.0133883870967737
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-9075 512 62.393432 0.4079886666666666 - 1.6319546666666664
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-9075 512 45.735847 0.078175 - 0.3127
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 80.363343 0.22575 - 0.903
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 72.259121 0.11525599999999998 - 0.46102399999999993
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 56.965983 0.03423233333333333 - 0.13692933333333332
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-9075 4096 51.774191 0.5019311290322581 - 16.06179612903226
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-9075 4096 36.154466 0.08403467741935483 - 2.6891096774193546
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 45.639921 0.48976935483870965 - 15.672619354838709
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 57.849112 0.16003048387096772 - 5.120975483870967
LFM2.5-1.2B-Instruct GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 46.243552 0.0413068064516129 - 1.3218178064516128

License

  • The license for the original implementation of LFM2.5-1.2B-Instruct 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