Nemotron 3 Super 120B

Nemotron 3 Super 120B is NVIDIA's flagship language model in the Nemotron family, featuring 120 billion parameters designed for sophisticated text processing and generation tasks. As NVIDIA's top-tier offering in this model line, it represents the company's entry into large-scale language modeling alongside their established GPU and AI infrastructure products. The model operates with a 262,144 token context window, enabling processing of lengthy documents and extended conversations. It focuses exclusively on text modalities and delivers measured performance with an output rate of 164.34 tokens per second and a time to first token of 744 milliseconds according to Artificial Analysis benchmarks. The model does not include tool calling capabilities, positioning it as a pure language processing solution. Nemotron 3 Super 120B serves users requiring substantial language understanding and generation capabilities within NVIDIA's ecosystem. While newer flagship models from other providers have emerged since its release, it remains NVIDIA's primary large language model offering for complex text-based applications where extended context and substantial model capacity are priorities.