MiniLM-v2: Optimized for Qualcomm Devices

All-MiniLM-L6-v2 maps sentences to a 384-dimensional dense vector space. Trained on 1B+ sentence pairs, it excels at semantic search, clustering, and sentence similarity tasks while being small enough to run on mobile devices.

This is based on the implementation of MiniLM-v2 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.

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
ONNX	float	Universal	QAIRT 2.45, ONNX Runtime 1.25.0	Download
ONNX	w8a8	Universal	QAIRT 2.45, ONNX Runtime 1.25.0	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a8	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download
TFLITE	w8a8	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit MiniLM-v2 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 MiniLM-v2 on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.text_generation

Model Stats:

• Model checkpoint: sentence-transformers/all-MiniLM-L6-v2
• Input resolution: 128 tokens
• Number of parameters: 22.7M
• Model size (float): 86.7 MB
• Embedding dimension: 384

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
MiniLM-v2	ONNX	float	Snapdragon® X2 Elite	0.781 ms	1 - 1 MB	NPU
MiniLM-v2	ONNX	float	Snapdragon® X Elite	1.751 ms	45 - 45 MB	NPU
MiniLM-v2	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.227 ms	0 - 104 MB	NPU
MiniLM-v2	ONNX	float	Snapdragon® 8 Gen 1 Mobile	2.547 ms	0 - 108 MB	NPU
MiniLM-v2	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.71 ms	0 - 3 MB	NPU
MiniLM-v2	ONNX	float	Qualcomm® QCS8450	2.547 ms	0 - 108 MB	NPU
MiniLM-v2	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	2.074 ms	0 - 3 MB	NPU
MiniLM-v2	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	0.7 ms	0 - 62 MB	NPU
MiniLM-v2	ONNX	float	Snapdragon® 8 Elite Mobile	0.864 ms	0 - 60 MB	NPU
MiniLM-v2	ONNX	float	Qualcomm® Dragonwing™ Q-8750	0.864 ms	0 - 60 MB	NPU
MiniLM-v2	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	1.751 ms	45 - 45 MB	NPU
MiniLM-v2	ONNX	w8a8	Snapdragon® X2 Elite	0.673 ms	1 - 1 MB	NPU
MiniLM-v2	ONNX	w8a8	Snapdragon® X Elite	1.783 ms	23 - 23 MB	NPU
MiniLM-v2	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	1.202 ms	0 - 85 MB	NPU
MiniLM-v2	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	1.973 ms	0 - 87 MB	NPU
MiniLM-v2	ONNX	w8a8	Qualcomm® Dragonwing™ QCS6490	5.522 ms	0 - 3 MB	NPU
MiniLM-v2	ONNX	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.704 ms	0 - 2 MB	NPU
MiniLM-v2	ONNX	w8a8	Qualcomm® QCS8450	1.973 ms	0 - 87 MB	NPU
MiniLM-v2	ONNX	w8a8	Snapdragon® 8 Elite Mobile	0.845 ms	0 - 57 MB	NPU
MiniLM-v2	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	1.939 ms	0 - 55 MB	NPU
MiniLM-v2	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.666 ms	0 - 62 MB	NPU
MiniLM-v2	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-9075	1.8 ms	0 - 3 MB	NPU
MiniLM-v2	ONNX	w8a8	Qualcomm® Dragonwing™ Q-6690	6.703 ms	0 - 61 MB	NPU
MiniLM-v2	ONNX	w8a8	Qualcomm® Dragonwing™ Q-7790	1.939 ms	0 - 55 MB	NPU
MiniLM-v2	ONNX	w8a8	Qualcomm® Dragonwing™ Q-8750	0.845 ms	0 - 57 MB	NPU
MiniLM-v2	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-X7181	1.783 ms	23 - 23 MB	NPU
MiniLM-v2	QNN_DLC	float	Snapdragon® X2 Elite	0.672 ms	1 - 1 MB	NPU
MiniLM-v2	QNN_DLC	float	Snapdragon® X Elite	1.3 ms	1 - 1 MB	NPU
MiniLM-v2	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	0.797 ms	0 - 100 MB	NPU
MiniLM-v2	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	1.888 ms	0 - 100 MB	NPU
MiniLM-v2	QNN_DLC	float	Qualcomm® QCS8275	3.554 ms	0 - 58 MB	NPU
MiniLM-v2	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.111 ms	0 - 20 MB	NPU
MiniLM-v2	QNN_DLC	float	Qualcomm® SA8775P	1.513 ms	0 - 60 MB	NPU
MiniLM-v2	QNN_DLC	float	Qualcomm® SA8650P	1.513 ms	0 - 60 MB	NPU
MiniLM-v2	QNN_DLC	float	Qualcomm® SA8255P	1.513 ms	0 - 60 MB	NPU
MiniLM-v2	QNN_DLC	float	Qualcomm® QCS8450	1.888 ms	0 - 100 MB	NPU
MiniLM-v2	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	1.389 ms	0 - 2 MB	NPU
MiniLM-v2	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.459 ms	0 - 61 MB	NPU
MiniLM-v2	QNN_DLC	float	Qualcomm® SA7255P	3.554 ms	0 - 58 MB	NPU
MiniLM-v2	QNN_DLC	float	Snapdragon® 8 Elite Mobile	0.563 ms	0 - 59 MB	NPU
MiniLM-v2	QNN_DLC	float	Qualcomm® SA8295P	2.293 ms	0 - 56 MB	NPU
MiniLM-v2	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	0.563 ms	0 - 59 MB	NPU
MiniLM-v2	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	1.3 ms	1 - 1 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Snapdragon® X2 Elite	0.8 ms	0 - 0 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Snapdragon® X Elite	1.903 ms	0 - 0 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	1.211 ms	3 - 72 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	1.944 ms	0 - 74 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS6490	4.196 ms	2 - 4 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® QCS8275	3.465 ms	0 - 48 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.727 ms	0 - 15 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® SA8775P	1.951 ms	0 - 51 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® SA8650P	1.951 ms	0 - 51 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® SA8255P	1.951 ms	0 - 51 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® QCS8450	1.944 ms	0 - 74 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	0.832 ms	0 - 48 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	1.701 ms	0 - 47 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.677 ms	0 - 50 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-9075	1.821 ms	0 - 2 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-6690	6.131 ms	0 - 50 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® SA7255P	3.465 ms	0 - 48 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® SA8295P	2.357 ms	0 - 46 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-7790	1.701 ms	0 - 47 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-8750	0.832 ms	0 - 48 MB	NPU
MiniLM-v2	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-X7181	1.903 ms	0 - 0 MB	NPU
MiniLM-v2	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	0.819 ms	0 - 106 MB	NPU
MiniLM-v2	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	1.966 ms	0 - 102 MB	NPU
MiniLM-v2	TFLITE	float	Qualcomm® QCS8275	3.676 ms	0 - 63 MB	NPU
MiniLM-v2	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.117 ms	0 - 2 MB	NPU
MiniLM-v2	TFLITE	float	Qualcomm® SA8775P	1.565 ms	0 - 64 MB	NPU
MiniLM-v2	TFLITE	float	Qualcomm® SA8650P	1.565 ms	0 - 64 MB	NPU
MiniLM-v2	TFLITE	float	Qualcomm® SA8255P	1.565 ms	0 - 64 MB	NPU
MiniLM-v2	TFLITE	float	Qualcomm® QCS8450	1.966 ms	0 - 102 MB	NPU
MiniLM-v2	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	1.408 ms	0 - 45 MB	NPU
MiniLM-v2	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.487 ms	0 - 58 MB	NPU
MiniLM-v2	TFLITE	float	Qualcomm® SA7255P	3.676 ms	0 - 63 MB	NPU
MiniLM-v2	TFLITE	float	Snapdragon® 8 Elite Mobile	0.597 ms	0 - 60 MB	NPU
MiniLM-v2	TFLITE	float	Qualcomm® SA8295P	2.361 ms	0 - 56 MB	NPU
MiniLM-v2	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	0.597 ms	0 - 60 MB	NPU
MiniLM-v2	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	1.668 ms	0 - 80 MB	NPU
MiniLM-v2	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	2.624 ms	0 - 80 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS6490	16.748 ms	0 - 31 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® QCS8275	4.624 ms	0 - 54 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	2.367 ms	0 - 4 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® SA8775P	2.653 ms	0 - 59 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® SA8650P	2.653 ms	0 - 59 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® SA8255P	2.653 ms	0 - 59 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® QCS8450	2.624 ms	0 - 80 MB	NPU
MiniLM-v2	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	1.136 ms	0 - 61 MB	NPU
MiniLM-v2	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	5.071 ms	0 - 35 MB	NPU
MiniLM-v2	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.947 ms	0 - 53 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® Dragonwing™ IQ-9075	2.51 ms	0 - 24 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-6690	11.97 ms	0 - 37 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® SA7255P	4.624 ms	0 - 54 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® SA8295P	3.209 ms	0 - 48 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-7790	5.071 ms	0 - 35 MB	NPU
MiniLM-v2	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-8750	1.136 ms	0 - 61 MB	NPU

License

• The license for the original implementation of MiniLM-v2 can be found here.

References

• Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
• 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.