v0.65.0

8fbbf04 verified 9 days ago

11.5 kB

---
library_name: pytorch
license: apache-2.0
tags:
- llm
- vlm
- generative_ai
- android
pipeline_tag: text-generation

---
![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/smolvlm2_2_2b_instruct/web-assets/model_demo.png)

# SmolVLM2-2.2B-Instruct: Optimized for Qualcomm Devices

SmolVLM2 is a lightweight vision-language model from Hugging Face capable of understanding text and images for tasks such as visual question answering, image captioning, and optical character recognition.

This is based on the implementation of SmolVLM2-2.2B-Instruct found [here](https://huggingface.co/HuggingFaceTB/SmolVLM2-2.2B-Instruct).
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/v0.65.0/src/qai_hub_models/models/smolvlm2_2_2b_instruct) 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.

## Deploying SmolVLM2-2.2B-Instruct on-device

Follow the [GenieX quickstart](https://geniex.aihub.qualcomm.com/en/get-started/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 | q8_0 | Universal |  | [Download](https://huggingface.co/ggml-org/SmolVLM2-2.2B-Instruct-GGUF/resolve/main/SmolVLM2-2.2B-Instruct-Q8_0.gguf)

For more device-specific assets and performance metrics, visit **[SmolVLM2-2.2B-Instruct on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/smolvlm2_2_2b_instruct)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.65.0/src/qai_hub_models/models/smolvlm2_2_2b_instruct) 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 [SmolVLM2-2.2B-Instruct on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.65.0/src/qai_hub_models/models/smolvlm2_2_2b_instruct) for usage instructions.

## Model Details

**Model Type:** Model_use_case.text_generation
**Model Stats:**
## Performance Summary
| Model | Runtime | Precision | Chipset | Context Length | Response Rate (tokens per second) | Time To First Token (range, seconds)
|---|---|---|---|---|---|---
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 512 | 29.436277 | 0.38540099999999994 - 1.5416039999999998
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 512 | 28.538412 | 0.24594099999999997 - 0.9837639999999999
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 512 | 28.889784 | 0.051866999999999996 - 0.20746799999999999
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 4096 | 11.47742 | 0.9924147096774193 - 31.757270709677417
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 4096 | 15.556558 | 0.4533808064516129 - 14.508185806451612
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 4096 | 14.064579 | 0.0729842258064516 - 2.335495225806451
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite For Galaxy Mobile | 512 | 28.134663 | 0.47438633333333335 - 1.8975453333333334
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite For Galaxy Mobile | 512 | 28.243021 | 0.2631319999999999 - 1.0525279999999997
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite For Galaxy Mobile | 512 | 24.911554 | 0.06054233333333333 - 0.24216933333333332
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite For Galaxy Mobile | 4096 | 11.020719 | 1.1051197096774195 - 35.36383070967742
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite For Galaxy Mobile | 4096 | 15.885515 | 0.5819834193548387 - 18.623469419354837
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® 8 Elite For Galaxy Mobile | 4096 | 14.225878 | 0.10105106451612902 - 3.233634064516129
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X2 Elite | 512 | 49.184781 | 0.13440833333333332 - 0.5376333333333333
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X2 Elite | 512 | 32.516632 | 0.11517866666666665 - 0.4607146666666666
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X2 Elite | 512 | 25.874832 | 0.050324666666666656 - 0.20129866666666663
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X2 Elite | 4096 | 24.564593 | 0.21470767741935484 - 6.870645677419355
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X2 Elite | 4096 | 18.539222 | 0.23938735483870965 - 7.660395354838709
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X2 Elite | 4096 | 13.754858 | 0.07732448387096774 - 2.4743834838709677
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X Elite | 512 | 20.396668 | 0.33751333333333333 - 1.3500533333333333
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X Elite | 512 | 32.472137 | 0.2929063333333333 - 1.1716253333333333
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X Elite | 512 | 21.951898 | 0.12050033333333332 - 0.4820013333333333
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X Elite | 4096 | 13.397349 | 0.5872530322580646 - 18.792097032258066
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X Elite | 4096 | 11.692214 | 0.565842064516129 - 18.106946064516126
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Snapdragon® X Elite | 4096 | 8.764435 | 0.1703770322580645 - 5.452065032258064
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-8275 | 512 | 8.077908 | 2.271742333333333 - 9.086969333333332
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-8275 | 512 | 5.158383 | 2.4926939999999997 - 9.970775999999999
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-8275 | 512 | 5.613499 | 0.21356999999999998 - 0.8542799999999999
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-8275 | 4096 | 2.684636 | 3.4063445806451615 - 109.00302658064517
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-8275 | 4096 | 3.218268 | 3.5844897419354838 - 114.70367174193548
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-8275 | 4096 | 5.0273 | 0.2295644516129032 - 7.3460624516129025
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-9075 | 512 | 24.016041 | 0.8226986666666667 - 3.2907946666666668
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-9075 | 512 | 13.361646 | 0.8900710000000001 - 3.5602840000000002
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-9075 | 512 | 7.807072 | 0.175873 - 0.703492
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-X7181 | 512 | 20.396668 | 0.33751333333333333 - 1.3500533333333333
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-X7181 | 512 | 32.472137 | 0.2929063333333333 - 1.1716253333333333
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-X7181 | 512 | 21.951898 | 0.12050033333333332 - 0.4820013333333333
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ Q-8750 | 512 | 28.134663 | 0.47438633333333335 - 1.8975453333333334
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ Q-8750 | 512 | 28.243021 | 0.2631319999999999 - 1.0525279999999997
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ Q-8750 | 512 | 24.911554 | 0.06054233333333333 - 0.24216933333333332
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-9075 | 4096 | 15.101235 | 1.1813366451612903 - 37.80277264516129
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-9075 | 4096 | 8.021626 | 1.3425965483870967 - 42.963089548387096
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-9075 | 4096 | 6.112163 | 0.19832819354838707 - 6.346502193548386
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-X7181 | 4096 | 13.397349 | 0.5872530322580646 - 18.792097032258066
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-X7181 | 4096 | 11.692214 | 0.565842064516129 - 18.106946064516126
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ IQ-X7181 | 4096 | 8.764435 | 0.1703770322580645 - 5.452065032258064
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ Q-8750 | 4096 | 11.020719 | 1.1051197096774195 - 35.36383070967742
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ Q-8750 | 4096 | 15.885515 | 0.5819834193548387 - 18.623469419354837
| SmolVLM2-2.2B-Instruct | GENIEX_LLAMACPP | q8_0 | Qualcomm® Dragonwing™ Q-8750 | 4096 | 14.225878 | 0.10105106451612902 - 3.233634064516129
## License
* The license for the original implementation of SmolVLM2-2.2B-Instruct can be found
  [here](https://www.apache.org/licenses/LICENSE-2.0).
## References
* [SmolVLM: Redefining small and efficient multimodal models](https://arxiv.org/abs/2504.05299)
* [Source Model Implementation](https://huggingface.co/HuggingFaceTB/SmolVLM2-2.2B-Instruct)
## 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).
## 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

	---
	library_name: pytorch
	license: apache-2.0
	tags:
	- llm
	- vlm
	- generative_ai
	- android
	pipeline_tag: text-generation

	---

	![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/smolvlm2_2_2b_instruct/web-assets/model_demo.png)

	# SmolVLM2-2.2B-Instruct: Optimized for Qualcomm Devices

	SmolVLM2 is a lightweight vision-language model from Hugging Face capable of understanding text and images for tasks such as visual question answering, image captioning, and optical character recognition.

	This is based on the implementation of SmolVLM2-2.2B-Instruct found [here](https://huggingface.co/HuggingFaceTB/SmolVLM2-2.2B-Instruct).
	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/v0.65.0/src/qai_hub_models/models/smolvlm2_2_2b_instruct) 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.

	## Deploying SmolVLM2-2.2B-Instruct on-device

	Follow the [GenieX quickstart](https://geniex.aihub.qualcomm.com/en/get-started/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 \| q8_0 \| Universal \| \| [Download](https://huggingface.co/ggml-org/SmolVLM2-2.2B-Instruct-GGUF/resolve/main/SmolVLM2-2.2B-Instruct-Q8_0.gguf)

	For more device-specific assets and performance metrics, visit [SmolVLM2-2.2B-Instruct on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/smolvlm2_2_2b_instruct).


	### Option 2: Export with Custom Configurations

	Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.65.0/src/qai_hub_models/models/smolvlm2_2_2b_instruct) 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 [SmolVLM2-2.2B-Instruct on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.65.0/src/qai_hub_models/models/smolvlm2_2_2b_instruct) for usage instructions.

	## Model Details

	Model Type: Model_use_case.text_generation

	Model Stats:


	## Performance Summary
	\| Model \| Runtime \| Precision \| Chipset \| Context Length \| Response Rate (tokens per second) \| Time To First Token (range, seconds)
	\|---\|---\|---\|---\|---\|---\|---
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite Gen 5 For Galaxy Mobile \| 512 \| 29.436277 \| 0.38540099999999994 - 1.5416039999999998
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite Gen 5 For Galaxy Mobile \| 512 \| 28.538412 \| 0.24594099999999997 - 0.9837639999999999
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite Gen 5 For Galaxy Mobile \| 512 \| 28.889784 \| 0.051866999999999996 - 0.20746799999999999
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite Gen 5 For Galaxy Mobile \| 4096 \| 11.47742 \| 0.9924147096774193 - 31.757270709677417
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite Gen 5 For Galaxy Mobile \| 4096 \| 15.556558 \| 0.4533808064516129 - 14.508185806451612
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite Gen 5 For Galaxy Mobile \| 4096 \| 14.064579 \| 0.0729842258064516 - 2.335495225806451
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite For Galaxy Mobile \| 512 \| 28.134663 \| 0.47438633333333335 - 1.8975453333333334
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite For Galaxy Mobile \| 512 \| 28.243021 \| 0.2631319999999999 - 1.0525279999999997
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite For Galaxy Mobile \| 512 \| 24.911554 \| 0.06054233333333333 - 0.24216933333333332
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite For Galaxy Mobile \| 4096 \| 11.020719 \| 1.1051197096774195 - 35.36383070967742
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite For Galaxy Mobile \| 4096 \| 15.885515 \| 0.5819834193548387 - 18.623469419354837
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® 8 Elite For Galaxy Mobile \| 4096 \| 14.225878 \| 0.10105106451612902 - 3.233634064516129
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X2 Elite \| 512 \| 49.184781 \| 0.13440833333333332 - 0.5376333333333333
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X2 Elite \| 512 \| 32.516632 \| 0.11517866666666665 - 0.4607146666666666
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X2 Elite \| 512 \| 25.874832 \| 0.050324666666666656 - 0.20129866666666663
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X2 Elite \| 4096 \| 24.564593 \| 0.21470767741935484 - 6.870645677419355
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X2 Elite \| 4096 \| 18.539222 \| 0.23938735483870965 - 7.660395354838709
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X2 Elite \| 4096 \| 13.754858 \| 0.07732448387096774 - 2.4743834838709677
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X Elite \| 512 \| 20.396668 \| 0.33751333333333333 - 1.3500533333333333
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X Elite \| 512 \| 32.472137 \| 0.2929063333333333 - 1.1716253333333333
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X Elite \| 512 \| 21.951898 \| 0.12050033333333332 - 0.4820013333333333
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X Elite \| 4096 \| 13.397349 \| 0.5872530322580646 - 18.792097032258066
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X Elite \| 4096 \| 11.692214 \| 0.565842064516129 - 18.106946064516126
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Snapdragon® X Elite \| 4096 \| 8.764435 \| 0.1703770322580645 - 5.452065032258064
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-8275 \| 512 \| 8.077908 \| 2.271742333333333 - 9.086969333333332
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-8275 \| 512 \| 5.158383 \| 2.4926939999999997 - 9.970775999999999
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-8275 \| 512 \| 5.613499 \| 0.21356999999999998 - 0.8542799999999999
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-8275 \| 4096 \| 2.684636 \| 3.4063445806451615 - 109.00302658064517
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-8275 \| 4096 \| 3.218268 \| 3.5844897419354838 - 114.70367174193548
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-8275 \| 4096 \| 5.0273 \| 0.2295644516129032 - 7.3460624516129025
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-9075 \| 512 \| 24.016041 \| 0.8226986666666667 - 3.2907946666666668
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-9075 \| 512 \| 13.361646 \| 0.8900710000000001 - 3.5602840000000002
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-9075 \| 512 \| 7.807072 \| 0.175873 - 0.703492
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-X7181 \| 512 \| 20.396668 \| 0.33751333333333333 - 1.3500533333333333
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-X7181 \| 512 \| 32.472137 \| 0.2929063333333333 - 1.1716253333333333
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-X7181 \| 512 \| 21.951898 \| 0.12050033333333332 - 0.4820013333333333
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ Q-8750 \| 512 \| 28.134663 \| 0.47438633333333335 - 1.8975453333333334
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ Q-8750 \| 512 \| 28.243021 \| 0.2631319999999999 - 1.0525279999999997
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ Q-8750 \| 512 \| 24.911554 \| 0.06054233333333333 - 0.24216933333333332
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-9075 \| 4096 \| 15.101235 \| 1.1813366451612903 - 37.80277264516129
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-9075 \| 4096 \| 8.021626 \| 1.3425965483870967 - 42.963089548387096
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-9075 \| 4096 \| 6.112163 \| 0.19832819354838707 - 6.346502193548386
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-X7181 \| 4096 \| 13.397349 \| 0.5872530322580646 - 18.792097032258066
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-X7181 \| 4096 \| 11.692214 \| 0.565842064516129 - 18.106946064516126
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ IQ-X7181 \| 4096 \| 8.764435 \| 0.1703770322580645 - 5.452065032258064
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ Q-8750 \| 4096 \| 11.020719 \| 1.1051197096774195 - 35.36383070967742
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ Q-8750 \| 4096 \| 15.885515 \| 0.5819834193548387 - 18.623469419354837
	\| SmolVLM2-2.2B-Instruct \| GENIEX_LLAMACPP \| q8_0 \| Qualcomm® Dragonwing™ Q-8750 \| 4096 \| 14.225878 \| 0.10105106451612902 - 3.233634064516129

	## License
	* The license for the original implementation of SmolVLM2-2.2B-Instruct can be found
	[here](https://www.apache.org/licenses/LICENSE-2.0).

	## References
	* [SmolVLM: Redefining small and efficient multimodal models](https://arxiv.org/abs/2504.05299)
	* [Source Model Implementation](https://huggingface.co/HuggingFaceTB/SmolVLM2-2.2B-Instruct)

	## 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).

	## 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