Qwen3-1.7B: Optimized for Qualcomm Devices

Qwen3 is a multilingual large language model series from Alibaba Cloud, excelling in language understanding, generation, coding, and mathematics.

This is based on the implementation of Qwen3-1.7B 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 Qwen3-1.7B 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 Qwen3-1.7B 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 Qwen3-1.7B on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.text_generation

Model Stats:

• Model architecture: Dense Transformer with Grouped Query Attention (GQA) and SwiGLU activation.
• 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.
• Response Rate: Rate of response generation after the first response token.

Performance Summary

Model	Runtime	Precision	Chipset	Context Length	Response Rate (tokens per second)	Time To First Token (range, seconds)
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	512	47.814629	0.40107475 - 1.604299
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	512	47.324852	0.43204575 - 1.728183
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	512	34.784811	0.06557025 - 0.262281
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	4096	27.781974	0.97570596875 - 31.222591
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	4096	22.642492	1.2193096250000002 - 39.017908000000006
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	4096	22.279084	0.11484521875 - 3.675047
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	512	43.700444	0.464084 - 1.856336
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	512	44.018186	0.46250775 - 1.850031
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	512	35.927586	0.09946899999999999 - 0.39787599999999995
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	4096	13.296329	1.0056319999999999 - 32.180223999999995
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	4096	13.132999	1.0708864375 - 34.268366
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	4096	21.232063	0.19318968749999998 - 6.1820699999999995
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	512	64.69927	0.11687925 - 0.467517
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	512	64.899408	0.11430625000000001 - 0.45722500000000005
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	512	38.419333	0.06460275 - 0.258411
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	4096	36.023298	0.23819934375 - 7.622379
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	4096	36.289295	0.2382465 - 7.623888
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	4096	22.904047	0.10138803125 - 3.244417
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	512	47.757915	0.181332 - 0.725328
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	512	45.855044	0.20409975 - 0.816399
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	512	27.030463	0.15535149999999998 - 0.6214059999999999
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	4096	25.55107	0.4618461875 - 14.779078
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	4096	25.258525	0.483645375 - 15.476652
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	4096	13.571355	0.24020603125 - 7.686593
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	4096	23.5	0.9741248100000001 - 31.171993999999998
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	4096	5.4	1.2736318409999998 - 40.756218999999994
Qwen3-1.7B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	4096	15.2	0.254321478 - 8.138287

License

• The license for the original implementation of Qwen3-1.7B can be found here.

References

• Qwen3 Technical Report
• Source Model Implementation

Community

• Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
• For questions or feedback please reach out to us.

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