SmolVLM Grows Smaller – Introducing the 256M & 500M Models!

The SmolVLM project has introduced two new model sizes: 256M and 500M parameters, expanding its range of lightweight visual-language models. These smaller models maintain competitive performance on standard benchmarks while significantly reducing computational requirements, making them more accessible for deployment in resource-constrained environments. This release is particularly relevant for practitioners seeking efficient solutions for integrating vision and language tasks without the overhead of larger models.