README.md

---
library_name: pytorch
license: other
tags:
- foundation
- android
pipeline_tag: image-segmentation

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/sam3/web-assets/model_demo.png)

# Segment-Anything-Model-3: Optimized for Qualcomm Devices

SAM3 (Segment Anything with Concepts) extends SAM2 with open-vocabulary segmentation, producing bounding boxes and masks for objects matching a natural-language prompt.

This is based on the implementation of Segment-Anything-Model-3 found [here](https://github.com/facebookresearch/sam3).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.62.1/src/qai_hub_models/models/sam3) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.

## Getting Started
Due to licensing restrictions, we cannot distribute pre-exported model assets for this model.
Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.62.1/src/qai_hub_models/models/sam3) 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

See our repository for [Segment-Anything-Model-3 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.62.1/src/qai_hub_models/models/sam3) for usage instructions.


## Model Details

**Model Type:** Model_use_case.semantic_segmentation

**Model Stats:**
- Input resolution: 1008x1008
- Model checkpoint: sam3
- Model size (SAM3Backbone) (float): 128 MB
- Model size (SAM3Transformer) (float): 23.7 MB
- Number of parameters (SAM3Backbone): 33.5M
- Number of parameters (SAM3Transformer): 6.22M

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| head | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 208.916 ms | 136 - 136 MB | NPU
| head | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 483.955 ms | 667 - 667 MB | NPU
| head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 381.819 ms | 155 - 168 MB | NPU
| head | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 490.846 ms | 106 - 216 MB | NPU
| head | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 501.25 ms | 106 - 215 MB | NPU
| head | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 483.955 ms | 667 - 667 MB | NPU
| head | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 276.292 ms | 132 - 144 MB | NPU
| head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Mobile | 276.292 ms | 132 - 144 MB | NPU
| head | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 236.91 ms | 131 - 143 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 209.654 ms | 106 - 106 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 483.87 ms | 107 - 107 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 385.86 ms | 106 - 119 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | 491.758 ms | 106 - 278 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | 516.497 ms | 97 - 107 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8650P | 516.497 ms | 97 - 107 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8255P | 516.497 ms | 97 - 107 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | 502.225 ms | 106 - 278 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-X7181 | 483.87 ms | 107 - 107 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ Q-8750 | 283.287 ms | 82 - 90 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | 1004.109 ms | 106 - 116 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 595.703 ms | 82 - 87 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Mobile | 283.287 ms | 82 - 90 MB | NPU
| head | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | 230.122 ms | 61 - 74 MB | NPU
| vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 1081.01 ms | 265 - 265 MB | NPU
| vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 2224.923 ms | 943 - 943 MB | NPU
| vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1744.17 ms | 31 - 41 MB | NPU
| vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 2529.726 ms | 114 - 129 MB | NPU
| vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 2458.695 ms | 54 - 68 MB | NPU
| vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 2224.923 ms | 943 - 943 MB | NPU
| vision_backbone | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1410.014 ms | 8 - 20 MB | NPU
| vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Mobile | 1410.014 ms | 8 - 20 MB | NPU
| vision_backbone | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1227.166 ms | 8 - 20 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 1092.428 ms | 12 - 12 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 2188.246 ms | 12 - 12 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 1732.551 ms | 12 - 26 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | 2528.326 ms | 12 - 132 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | 2520.912 ms | 1 - 11 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8650P | 2520.912 ms | 1 - 11 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8255P | 2520.912 ms | 1 - 11 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | 2453.533 ms | 12 - 131 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-X7181 | 2188.246 ms | 12 - 12 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ Q-8750 | 1406.407 ms | 0 - 13 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | 6422.916 ms | 7 - 16 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 2784.464 ms | 1 - 11 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Mobile | 1406.407 ms | 0 - 13 MB | NPU
| vision_backbone | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | 1249.631 ms | 11 - 21 MB | NPU

## License
* The license for the original implementation of Segment-Anything-Model-3 can be found
  [here](https://github.com/facebookresearch/sam3/blob/main/LICENSE).

## References
* [SAM 3: Segment Anything with Concepts](https://arxiv.org/abs/2511.16719)
* [Source Model Implementation](https://github.com/facebookresearch/sam3)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).