| Property | Value |
|---|---|
| Parameter Count | 1.0B activated parameters |
| Model Type | Mixture-of-Experts Vision-Language Model |
| License | MIT License (code), DeepSeek Model License (model) |
| Paper | arXiv:2412.10302 |
What is deepseek-vl2-tiny?
DeepSeek-VL2-Tiny is part of the advanced DeepSeek-VL2 series, representing a significant evolution in vision-language models. Built on DeepSeekMoE-3B architecture, it offers 1.0B activated parameters, making it an efficient choice for multimodal understanding tasks while maintaining competitive performance.
Implementation Details
The model is implemented using a Mixture-of-Experts architecture, optimized for both computational efficiency and performance. It supports dynamic tiling strategy for up to 2 images and implements direct padding for 3 or more images at 384x384 resolution.
- Built on DeepSeekMoE-3B architecture
- Supports temperature-controlled sampling (recommended T ≤ 0.7)
- Implements efficient image handling strategies
- Python 3.8+ compatible
Core Capabilities
- Visual Question Answering
- Optical Character Recognition
- Document/Table/Chart Understanding
- Visual Grounding
- Multimodal Conversation
Frequently Asked Questions
Q: What makes this model unique?
DeepSeek-VL2-Tiny achieves competitive performance with fewer activated parameters through its innovative MoE architecture, making it particularly efficient for deployment while maintaining high-quality visual understanding capabilities.
Q: What are the recommended use cases?
The model excels in visual question answering, OCR tasks, document understanding, and visual grounding applications. It's particularly suitable for scenarios requiring efficient multimodal understanding with limited computational resources.