NVIDIA's open multimodal hybrid MoE model (120B total, 12B active) that reasons over text, images, documents and GUI screenshots, with built-in tool calling. Tuned for document understanding, OCR, charts and agentic workloads.
NVIDIA's open multimodal hybrid MoE model (120B total, 12B active) that reasons over text, images, documents and GUI screenshots, with built-in tool calling. Tuned for document understanding, OCR, charts and agentic workloads.