Qwen3.5-0.8B: Optimized for Qualcomm Devices

Qwen3.5 is the latest multilingual language model series from Alibaba Cloud with improved reasoning and instruction-following capabilities over Qwen3.

This is based on the implementation of Qwen3.5-0.8B 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 Qwen3.5-0.8B 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 Qwen3.5-0.8B 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 Qwen3.5-0.8B on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.text_generation

Model Stats:

Model architecture: Transformer with GQA and SwiGLU
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)
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	512	62.131097	0.23427574999999998 - 0.9371029999999999
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	512	50.763998	0.27429675 - 1.097187
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	512	35.869292	0.11880325 - 0.475213
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	4096	59.414176	0.36005078125 - 11.521625
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	4096	56.679703	0.39175446875 - 12.536143
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Gen 5 Mobile	4096	41.909392	0.11803384375 - 3.777083
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	512	72.950102	0.26924750000000003 - 1.0769900000000001
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	512	62.441461	0.26454675 - 1.058187
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	512	37.52486	0.1532055 - 0.612822
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	4096	49.751244	0.42967528125 - 13.749609
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	4096	48.911714	0.43504246875 - 13.921359
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® 8 Elite Mobile	4096	35.893754	0.1591485625 - 5.092754
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	512	88.774407	0.09157649999999999 - 0.36630599999999996
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	512	87.673154	0.0942735 - 0.377094
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	512	40.640275	0.1113045 - 0.445218
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	4096	81.095293	0.12939821875000002 - 4.1407430000000005
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	4096	80.304757	0.14290928125 - 4.573097
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X2 Elite	4096	39.613582	0.11075478124999999 - 3.5441529999999997
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	512	64.713531	0.12653075 - 0.506123
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	512	63.044187	0.12997225 - 0.519889
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	512	29.988464	0.2527895 - 1.011158
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	4096	51.537779	0.21393696875 - 6.845983
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	4096	48.821458	0.23768365625 - 7.605877
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Snapdragon® X Elite	4096	28.38794	0.25720621875 - 8.230599
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	512	49.130392	0.36298 - 1.45192
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	512	50.150451	0.36064100000000004 - 1.4425640000000002
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	512	14.985471	0.33844025 - 1.353761
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	4096	38.271652	0.46651725 - 14.928552
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	4096	39.346842	0.46202675 - 14.784856
Qwen3.5-0.8B	GENIEX_LLAMACPP	q4_0	Qualcomm® QCS9075	4096	14.03834	0.33502103125000005 - 10.720673000000001

License

The license for the original implementation of Qwen3.5-0.8B can be found here.

References

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

Downloads last month: -; Downloads are not tracked for this model. How to track