NVIDIA Nemotron Nano 12B v2 VL (free)

NVIDIA Nemotron Nano 12B v2 VL is an open 12-billion-parameter vision-language model built on a hybrid Mamba-Transformer architecture, designed for multimodal reasoning over documents, images, tables, charts, and video. It is optimized for high-throughput, low-latency inference on long-context multimodal workloads including OCR, document intelligence, and agentic pipelines. Source: supabase.