Gemma-SEA-LION-v4-4B-VL
Gemma-SEA-LION-v4-4B-VL adapts Gemma 3 4B IT for regional image and text tasks. It uses about 6.7 million instruction-text pairs. AI Singapore adds tool-calling data and examines visual parsing in Thai, Chinese and English.
How this release works
The model retains the image and text inputs of its Gemma parent. The regional language list includes Indonesian, Vietnamese, Thai, Filipino, Tamil, Burmese and Malay. The card reports separate tool-calling and visual-parsing work, which helps distinguish it from a text-only regional adaptation.Gemma-SEA-LION-v4-4B-VL official model card
How to assess it for your work
Practical guidance. Use a small set of real forms to check whether this compact release can extract the required fields. Include rotated text, low-resolution scans and mixed scripts. Review exact names and numbers separately from prose. A fluent explanation can still contain an incorrect field value.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | aisingapore/Gemma-SEA-LION-v4-4B-VL |
| Base model | google/gemma-3-4b-itGemma-SEA-LION-v4-4B-VL official model card |
| Context stated by the card | 128K tokensGemma-SEA-LION-v4-4B-VL official model card |
| Model family size | 4BGemma-SEA-LION-v4-4B-VL official model card |
| License declared by the model card | MITGemma-SEA-LION-v4-4B-VL official model card |
Related models
- Gemma-SEA-LION-v4-27B-VL
AI Singapore · SEA-LION v4
A 27B Gemma-based vision-language model trained on regional image-text instructions with SEACrowd.
Launch date not confirmedDownloadable - Gemma-SEA-LION-v4.5-E2B-IT
AI Singapore · SEA-LION v4.5
A Gemma 4 E2B adaptation for regional languages with text, image, video and audio inputs.
Launch date not confirmedDownloadable - Qwen-SEA-LION-v4-4B-VL
AI Singapore · SEA-LION v4
A 4B Qwen3-VL adaptation for regional image and text tasks with tool-use support.
Launch date not confirmedDownloadable