Qwen3-0.6B: Optimized for Qualcomm Devices

The Qwen3-0.6B is a state-of-the-art multilingual language model with 0.6 billion parameters, excelling in language understanding, generation, coding, and mathematics.

This is based on the implementation of Qwen3-0.6B 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 Qwen3-0.6B 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 Qwen3-0.6B 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 Qwen3-0.6B on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.text_generation

Model Stats:

• 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).
• 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)
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	512	109.93192	0.177354 - 0.709416
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	512	107.662002	0.1901215 - 0.760486
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	512	61.618566	0.041533499999999994 - 0.16613399999999998
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	4096	18.024064	0.71038396875 - 22.732287
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	4096	19.39173	0.8699334375 - 27.83787
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	4096	30.405844	0.07879000000000001 - 2.5212800000000004
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	512	101.447851	0.2191885 - 0.876754
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	512	101.683821	0.221036 - 0.884144
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	512	57.350625	0.06529425000000001 - 0.26117700000000005
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	4096	39.9992	0.75358753125 - 24.114801
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	4096	26.242931	0.7658371875000001 - 24.506790000000002
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	4096	28.394174	0.161257 - 5.160224
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	512	137.015188	0.05469325 - 0.218773
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	512	139.697424	0.055945 - 0.22378
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	512	53.395628	0.04178275 - 0.167131
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	4096	58.593223	0.17867675 - 5.717656
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	4096	58.535613	0.17892090625 - 5.725469
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	4096	27.68702	0.07636262499999999 - 2.4436039999999997
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	512	70.216514	0.13812225 - 0.552489
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	512	72.524369	0.19713 - 0.78852
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	512	43.02012	0.10337175000000001 - 0.41348700000000005
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	4096	29.268937	0.46896540625 - 15.006893
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	4096	31.137128	0.394078875 - 12.610524
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	4096	16.83887	0.18681084375 - 5.977947
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	512	80.733005	0.26653925 - 1.066157
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	512	74.240385	0.26632825000000004 - 1.0653130000000002
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	512	34.673153	0.10991074999999999 - 0.43964299999999995
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	4096	30.105746	0.70822534375 - 22.663211
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	4096	29.97025	0.70882696875 - 22.682463
Qwen3-0.6B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	4096	18.282076	0.1902083125 - 6.086666

License

• The license for the original implementation of Qwen3-0.6B 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