Laya Vision (SmolVLM-256M)

Laya Vision (SmolVLM-256M) is a lightweight open-source AI model designed for calibrated, typed decisions about images plus optional text input, answering choice, score, and yes/no probability questions in one forward pass. It is aimed at enabling visual question answering with a small vision-language model and usable via a Python API. The project is experimental and independent, hosted on GitHub and Hugging Face, suitable for developers and researchers focusing on AI vision capabilities.