AI applications are only as trustworthy as the safeguards behind them. As enterprises deploy increasingly capable multimodal AI systems, the need for scalable, customizable, and multilingual safety moderation has become critical.
That’s why Vultr is providing Day Zero support for NVIDIA Nemotron 3.5 Content Safety, a new small language model (SLM) designed to help developers and enterprises moderate AI inputs and outputs across text and images.
Built on Google’s Gemma-3-4B-it foundation model and fine-tuned by NVIDIA, Nemotron 3.5 Content Safety combines multimodal moderation across text, images, and custom policies with customizable policy reasoning to support a wide range of AI safety and governance use cases. It helps organizations deploy AI systems more safely while maintaining control over governance requirements.
What is NVIDIA Nemotron 3.5 Content Safety?
NVIDIA Nemotron 3.5 Content Safety is a compact 4B-parameter multimodal safety moderation model that evaluates prompts, responses, and images to determine whether content is safe or unsafe.
The model extends the capabilities of the original Nemotron 3 Content Safety model by adding support for:
- 23 safety categories based on Aegis v2 taxonomy
- Custom safety policy enforcement with reasoning-based explanations
- Reasoning-based moderation workflows
- Multilingual content moderation for 12 languages out of the box
- Multimodal moderation for text, images, and text-plus-images
- Domain-specific governance controls
- Up to 128K-token context window
The model supports both standard taxonomy classification and custom-policy reasoning modes, giving developers the flexibility to align moderation behavior with internal compliance, security, or industry-specific requirements.
Key capabilities
Nemotron 3.5 Content Safety is designed for modern AI applications that require scalable, real-time moderation across multiple modalities and languages.
Key capabilities include:
Multimodal moderation
The model can analyze both text and images within a single moderation workflow. Developers can evaluate user prompts, generated responses, and visual inputs together to detect unsafe or policy-violating content.
Custom policy reasoning
Unlike traditional guard models limited to fixed safety taxonomies, Nemotron 3.5 Content Safety supports custom-policy evaluation. Developers can define their own moderation rules and governance criteria while the model produces concise reasoning traces before classification.
Multilingual support
The model supports 12 languages out of the box, including English, Spanish, Mandarin, German, French, Hindi, Japanese, Arabic, and Thai, enabling global moderation workflows across enterprise AI deployments.
Long-context processing
With support for up to 128K context length, the model can evaluate large conversations, documents, and extended multimodal interactions without sacrificing moderation consistency.
Optimized for NVIDIA infrastructure
Nemotron 3.5 Content Safety is optimized for NVIDIA GPU-accelerated systems, including NVIDIA H100, and NVIDIA A100.
Built for enterprise AI governance
As organizations move AI systems into production, content safety and policy enforcement become foundational requirements.
Unlike traditional moderation models limited to fixed taxonomies, Nemotron 3.5 Content Safety allows organizations to evaluate content against custom policies and domain-specific governance requirements.
The model helps enterprises:
- Moderate prompts and responses in real time
- Reduce harmful or policy-violating outputs
- Support multilingual AI deployments
- Improve safety across multimodal applications
- Strengthen compliance, governance, and trust in AI systems
The model is designed for commercial use and integrates with popular inference frameworks, including Hugging Face Transformers, vLLM, and SGLang.
Deploy NVIDIA Nemotron 3.5 Content Safety on Vultr
Vultr provides the high-performance cloud GPU infrastructure needed to deploy and scale modern AI workloads globally.
Developers can use Vultr Cloud GPUs powered by NVIDIA to run Nemotron 3.5 Content Safety, offering enterprise-grade performance, scalability, and operational flexibility.
With Vultr, organizations can:
- Deploy globally across Vultr cloud regions
- Access high-performance NVIDIA GPUs on demand
- Scale inference workloads efficiently
- Support production AI moderation systems
- Accelerate AI application deployment
Get started
Want to start building with NVIDIA Nemotron 3.5 Content Safety? Get started today with the "nvidia/Nemotron-3.5-Content-Safety" model name.