NVIDIA Lists Nemotron 3.5 Content Safety: $0.20/M Tokens, 131k Context

What NVIDIA Just Listed

NVIDIA added Nemotron 3.5 Content Safety to its model roster on 2026-06-04. It's an open-source, multimodal safety classifier that accepts text and image inputs and returns text outputs — squarely aimed at developers who need to screen content at scale.

Pricing & Context Window

Both input and output are priced at $0.2 per 1M tokens, which is a flat, symmetric rate. That's straightforward to budget for: a 10M-token moderation run costs $2.00 in, $2.00 out, $4.00 total. The 131k context window is generous for a safety model — large enough to evaluate lengthy documents, conversation threads, or image-heavy prompts in a single pass rather than chunking.

Who Should Care

• High-volume platforms (social, UGC, marketplaces) will find the flat $0.2/$0.2 structure predictable. Compare this against incumbent moderation APIs on our site before committing.
• Multimodal pipelines benefit directly: most safety models are text-only. Native image support removes a preprocessing step.
• Open-source buyers get the model weights, which matters for teams with data-residency requirements or fine-tuning plans.

Caveats

Context on benchmark performance and false-positive rates isn't available yet — NVIDIA hasn't published evaluation numbers alongside the listing. Until independent benchmarks appear, treat accuracy claims with caution and run your own evals on domain-specific content.

For live pricing and spec updates, see Nemotron 3.5 Content Safety — live specs & price history.

Ezra, Scout AI Team