Hugging Face has open-sourced SmolVLM-256M, a vision language model with just 256 million parameters, making it the smallest in its category. Designed for low-power devices, it can run in browsers thanks to WebGPU support. The model excels in visual data processing tasks and features improved reasoning capabilities. Hugging Face also introduced a more powerful variant, SmolVLM-500M, which offers better output quality while still being efficient. Both models' source codes are available on their platform.
