NVIDIA: Nemotron Nano 12B 2 VL (free)

Nemotron Nano 12B 2 VL is a multimodal model from NVIDIA that accepts text, images, and video as input, with a context window of 128,000 tokens and a matching 128,000-token completion limit. It supports tool use and reasoning, which makes it applicable to agentic workflows and multi-step tasks. Structured output support is not confirmed. The model is currently free to use with no input or output costs. The free pricing makes it worth shortlisting for developers who want to experiment with video-capable multimodal pipelines without committing a budget. That said, there is no independent benchmark coverage available yet, so its relative performance against comparable models is unproven. Teams making a production decision should treat it as a candidate to test directly rather than one with an established track record, and should weigh the cost advantage against that uncertainty.