Hugging Face's logo

qualcomm
/
Qwen3-4B

Text Generation
PyTorch
llm
generative_ai
android
Model card Files
xet
 Community

Qwen3-4B: Optimized for Qualcomm Devices

The Qwen3-4B is a state-of-the-art multilingual base language model with 4 billion parameters, excelling in language understanding, generation, coding, and mathematics.

This is based on the implementation of Qwen3-4B 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 Qwen3-4B on-device

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

See the LLM-on-Genie tutorial to run with the Genie runtime. Note: Genie support will be deprecated soon.

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
GENIE w4a16 Snapdragon® 8 Elite Gen 5 Mobile QAIRT 2.45 Download
GENIE w4a16 Snapdragon® 8 Elite Mobile QAIRT 2.45 Download
GENIE w4a16 Snapdragon® X2 Elite QAIRT 2.45 Download
GENIE w4a16 Snapdragon® X Elite QAIRT 2.45 Download
GENIE w4a16 Qualcomm® Dragonwing™ IQ-8275 QAIRT 2.45 Download
GENIE w4a16 Qualcomm® SA8775P QAIRT 2.45 Download
GENIE w4a16 Qualcomm® Dragonwing™ IQ-9075 QAIRT 2.45 Download
GENIEX_LLAMACPP q4_0 Universal Download
GENIEX_QAIRT w4a16 Snapdragon® 8 Elite Gen 5 Mobile QAIRT 2.45 Download
GENIEX_QAIRT w4a16 Snapdragon® 8 Elite Mobile QAIRT 2.45 Download
GENIEX_QAIRT w4a16 Snapdragon® X2 Elite QAIRT 2.45 Download
GENIEX_QAIRT w4a16 Snapdragon® X Elite QAIRT 2.45 Download
GENIEX_QAIRT w4a16 Qualcomm® Dragonwing™ IQ-8275 QAIRT 2.45 Download
GENIEX_QAIRT w4a16 Qualcomm® Dragonwing™ IQ-9075 QAIRT 2.45 Download

For more device-specific assets and performance metrics, visit Qwen3-4B 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 Qwen3-4B on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.text_generation

Model Stats:

  • Response Rate: Rate of response generation after the first response token. Measured on a short prompt with a long response (with thinking); may slow down when using longer context lengths.
  • Supported languages: 100+ languages and dialects.
  • 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. The lower bound is for a short prompt (up to 128 tokens, i.e., one iteration of the prompt processor) and the upper bound is for a prompt using the full context length (4096 tokens).

Performance Summary

Model Runtime Precision Chipset Context Length Response Rate (tokens per second) Time To First Token (range, seconds)
Qwen3-4B GENIE w4a16 Snapdragon® 8 Elite Gen 5 Mobile 4096 29.241936683654785 0.051795499999999994 - 1.6574559999999998
Qwen3-4B GENIE w4a16 Snapdragon® 8 Elite Mobile 4096 24.391611099243164 0.06899839999999999 - 2.2079487999999996
Qwen3-4B GENIE w4a16 Snapdragon® X2 Elite 4096 30.268534088134764 0.06770319999999999 - 2.1665023999999997
Qwen3-4B GENIE w4a16 Snapdragon® X Elite 4096 6.564910411834717 0.10575100000000001 - 3.3840320000000004
Qwen3-4B GENIE w4a16 Qualcomm® Dragonwing™ IQ-9075 4096 17.71731367111206 0.10616810000000002 - 3.3973792000000005
Qwen3-4B GENIE w4a16 Qualcomm® Dragonwing™ IQ-X7181 4096 6.564910411834717 0.10575100000000001 - 3.3840320000000004
Qwen3-4B GENIE w4a16 Qualcomm® Dragonwing™ Q-8750 4096 24.391611099243164 0.06899839999999999 - 2.2079487999999996
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Mobile 512 22.5241 0.9030943333333332 - 3.612377333333333
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Mobile 512 19.213669 0.5367126666666666 - 2.1468506666666665
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Mobile 512 15.193501 0.128523 - 0.514092
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Mobile 4096 6.217811 2.6577142903225806 - 85.04685729032258
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Mobile 4096 7.210687 1.896787741935484 - 60.697207741935486
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® 8 Elite Mobile 4096 8.564574 0.1688638064516129 - 5.4036418064516125
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 33.741473 0.27790575 - 1.111623
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 33.490468 0.27597075 - 1.103883
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 512 20.973144 0.16287725 - 0.651509
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 20.323515 0.5701374375 - 18.244398
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 20.262276 0.57165421875 - 18.292935
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X2 Elite 4096 14.386503 0.2386385625 - 7.636434
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 512 26.208515 1.0620043333333333 - 4.248017333333333
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 512 21.600284 0.5379783333333332 - 2.151913333333333
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 512 11.405661 0.2611326666666666 - 1.0445306666666665
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 4096 12.76741 1.1816560000000003 - 37.81299200000001
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 4096 12.873649 1.3788717419354837 - 44.12389574193548
Qwen3-4B GENIEX_LLAMACPP q4_0 Snapdragon® X Elite 4096 5.462502 0.3414624838709677 - 10.926799483870967
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 5.81761 3.968277666666667 - 15.873110666666667
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 2.551969 5.7568069999999985 - 23.027227999999994
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 512 2.826793 0.4572486666666667 - 1.8289946666666668
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 1.11109 7.169169806451612 - 229.4134338064516
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 1.94841 8.773376419354838 - 280.7480454193548
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-8275 4096 2.259218 0.45057729032258065 - 14.41847329032258
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 512 26.208515 1.0620043333333333 - 4.248017333333333
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 512 21.600284 0.5379783333333332 - 2.151913333333333
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 512 11.405661 0.2611326666666666 - 1.0445306666666665
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 22.5241 0.9030943333333332 - 3.612377333333333
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 19.213669 0.5367126666666666 - 2.1468506666666665
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 512 15.193501 0.128523 - 0.514092
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 4096 12.76741 1.1816560000000003 - 37.81299200000001
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 4096 12.873649 1.3788717419354837 - 44.12389574193548
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ IQ-X7181 4096 5.462502 0.3414624838709677 - 10.926799483870967
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 6.217811 2.6577142903225806 - 85.04685729032258
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 7.210687 1.896787741935484 - 60.697207741935486
Qwen3-4B GENIEX_LLAMACPP q4_0 Qualcomm® Dragonwing™ Q-8750 4096 8.564574 0.1688638064516129 - 5.4036418064516125
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® 8 Elite Gen 5 Mobile 512 25.869943 0.061495666666666664 - 0.24598266666666666
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® 8 Elite Gen 5 Mobile 1024 25.327922 0.06677628571428572 - 0.5342102857142857
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® 8 Elite Gen 5 Mobile 2048 24.325005 0.07389533333333334 - 1.1823253333333335
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® 8 Elite Gen 5 Mobile 3072 22.356903 0.07948343478260868 - 1.9076024347826086
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® 8 Elite Gen 5 Mobile 4096 20.597869 0.08747154838709678 - 2.799089548387097
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® 8 Elite Mobile 4096 26.796653 0.05923 - 1.89536
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® X2 Elite 512 36.232202 0.05549066666666667 - 0.22196266666666667
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® X2 Elite 1024 35.70287 0.065999 - 0.527992
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® X2 Elite 2048 33.150222 0.06719386666666667 - 1.0751018666666667
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® X2 Elite 3072 30.538941 0.07345560869565218 - 1.762934608695652
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® X2 Elite 4096 26.975355 0.08009390322580645 - 2.5630049032258064
Qwen3-4B GENIEX_QAIRT w4a16 Snapdragon® X Elite 4096 21.220268 0.09903100000000001 - 3.1689920000000003
Qwen3-4B GENIEX_QAIRT w4a16 Qualcomm® Dragonwing™ IQ-8275 4096 16.769441 0.098417 - 3.149344
Qwen3-4B GENIEX_QAIRT w4a16 Qualcomm® Dragonwing™ IQ-9075 512 17.876586 0.09332866666666666 - 0.37331466666666663
Qwen3-4B GENIEX_QAIRT w4a16 Qualcomm® Dragonwing™ IQ-9075 1024 15.701198 0.10103628571428572 - 0.8082902857142857
Qwen3-4B GENIEX_QAIRT w4a16 Qualcomm® Dragonwing™ IQ-9075 2048 14.324252 0.11698886666666666 - 1.8718218666666666
Qwen3-4B GENIEX_QAIRT w4a16 Qualcomm® Dragonwing™ IQ-9075 3072 13.364087 0.1298216956521739 - 3.1157206956521737
Qwen3-4B GENIEX_QAIRT w4a16 Qualcomm® Dragonwing™ IQ-9075 4096 12.04256 0.1435604193548387 - 4.593933419354839
Qwen3-4B GENIEX_QAIRT w4a16 Qualcomm® Dragonwing™ IQ-X7181 4096 21.220268 0.09903100000000001 - 3.1689920000000003
Qwen3-4B GENIEX_QAIRT w4a16 Qualcomm® Dragonwing™ Q-8750 4096 26.796653 0.05923 - 1.89536

License

  • The license for the original implementation of Qwen3-4B 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/Qwen3-4B