NVIDIA Nemotron-3-Super-120B with Stained Glass Output Protection

Protopia AI's Stained Glass Transform (SGT) is a groundbreaking solution for organizations looking to leverage sensitive data in AI applications with robust security and flexibility. SGT transforms data into unintelligible forms, safeguarding against breaches and minimizing exposure risks without sacrificing fidelity. By enabling full data richness, SGT boosts value derived from AI models. This listing pairs SGT input protection with Stained Glass Output Protection: every generated token is returned tokenwise-encrypted (x25519 key exchange + AES-GCM), so the model's output is protected end to end, not just its input. With its low compute overhead, SGT operates efficiently on commodity hardware, freeing up powerful GPUs for critical tasks and offering greater flexibility across on-prem, multi-tenant, and edge environments. SGT's transformation process is fast, with minimal latency, integrating seamlessly into existing AI pipelines to maximize performance and throughput. The underlying NVIDIA-Nemotron-3-Super-120B-A12B model was pre-trained on over 25 trillion tokens of crawled and synthetic text spanning code, math, science, and general knowledge across 20 languages, then instruction-tuned for tool calling, structured outputs, and long-context retrieval. NVIDIA reports strong benchmark results for the base model, including 86.01 on MMLU and 90.67 on GSM8K (8-shot); this listing serves an FP8-quantized variant for reduced memory footprint and inference latency, which may show minor accuracy variation from the published BF16 figures. Tensor parallelism auto-sizes to the deployed instance's GPU count (TP=2 on g7e.12xlarge up to TP=8 on g7e.48xlarge), verified serving prompt_embeds correctly across all supported real-time instance sizes. For custom quotes and offers, please reach out at contact@protopia.ai