NVIDIA: Nemotron 3.5 Content Safety (free)

Nemotron 3.5 Content Safety is a text-and-image model from NVIDIA built specifically for content moderation and safety classification tasks. It accepts inputs up to 128,000 tokens and can return completions up to 8,192 tokens. The model supports reasoning but does not support tool calls or structured output, which limits its fit for agentic pipelines or schema-driven workflows. The model is free at both input and output, making the cost argument straightforward for teams that need to screen large volumes of text or image content without budget exposure. That said, there is no independent benchmark coverage available yet, so its performance relative to competing safety classifiers is unproven on third-party evaluations. Teams willing to run their own red-teaming and validation, or those using it to augment rather than replace existing moderation layers, are best positioned to shortlist it.