Nvidia: Nemotron 3 Super
NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model, activating just 12B parameters for maximum compute efficiency and accuracy in complex multi-agent applications. Built on a hybrid Mamba-Transformer... 120B MoE with only 12B active params for high efficiency 1M token context window for very long tasks Open-weight hybrid Mamba architecture Strong multi-agent application performance Newer architecture with less ecosystem support MoE routing can be inconsistent on edge cases Benchmark coverage still growing Complex multi-agent AI systems Long-context reasoning and analysis Efficient high-volume inference Enterprise agentic deployments
Specifications
| Lab | Nvidia |
|---|---|
| Context window | 1,000,000 |
| Input price | $0.09/1M |
| Output price | $0.45/1M |
| Release | 2026-03-01 00:00:00 |
Strengths
- 120B MoE with only 12B active params for high efficiency
- 1M token context window for very long tasks
- Open-weight hybrid Mamba architecture
- Strong multi-agent application performance
Weaknesses
- Newer architecture with less ecosystem support
- MoE routing can be inconsistent on edge cases
- Benchmark coverage still growing
Best for
- Complex multi-agent AI systems
- Long-context reasoning and analysis
- Efficient high-volume inference
- Enterprise agentic deployments
In Depth: Nemotron 3 Super
Summary
Nemotron 3 Super is an AI model from Nvidia.
Released 2026-03-01 00:00:00. It currently appears in the Overall category on LMRank. It supports Text input and produces Text output, with a context window of 1M tokens. Input pricing is $0.09 per 1M tokens and output is $0.45 per 1M tokens on OpenRouter.