Hugging Face, an open AI platform provider, has released SmolVLM, a compact vision-language AI model that processes images and text.

SmolVLM uses an image compression system that processes 384 × 384 image patches using 81 visual tokens. The model operates with 5.02 GB of GPU RAM compared to competitors that require 10–13 GB and can handle image and video analysis.

The company claims that SmolVLM can help businesses implement AI vision systems at lower costs while maintaining performance. The model's design could reduce computational resource requirements and make advanced vision-language capabilities accessible to companies with limited resources.

Analyst QuickTake: This follows the company’s earlier launches of SmolLM and SmolLM2 compact language models this year. As a small model that can effectively handle vision and language tasks , it challenges the prevailing notion that larger models are always better. The model's ability to achieve competitive performance with substantially fewer resources represents a crucial step toward democratizing AI technology.