FW-Nemotron-3-Super-120B-A12B-BF16

Models available for use with Fireworks on Foundry deliver optimized, best-in-class performance on the Fireworks Inference Cloud. Fireworks on Foundry is a Non-Microsoft Product. The following terms apply to a Customer's use of Fireworks on Foundry: When you use Fireworks on Foundry, data is shared between Microsoft and Fireworks AI, Customer Data will be sent outside of Microsoft systems, Customer Data will not be processed pursuant to any Foundry data residency documentation, and different compliance and data handling rules will apply. See Trust Center - Fireworks AI for details. Customers are responsible for evaluating whether data sharing between Microsoft and Fireworks is appropriate for their organization's compliance requirements.

Nemotron Super 120B (NVIDIA-Nemotron-3-Super-120B-A12B-BF16) is a large language model trained by NVIDIA for agentic workflows, long-context reasoning, and tool use. It employs a hybrid Latent Mixture-of-Experts (LatentMoE) architecture interleaving Mamba-2 state-space layers and Transformer attention layers, with 120B total parameters and 12B active parameters per token. The model supports a configurable reasoning mode and a native 262,144-token context window, with multilingual coverage across English, French, German, Italian, Japanese, Spanish, and Chinese.

• Hybrid Mamba-2 + LatentMoE architecture with select Transformer attention layers
• 120B total parameters with 12B active parameters per token
• 262,144-token native context window
• Configurable reasoning mode for agentic, multi-turn tasks
• Multilingual support (English, French, German, Italian, Japanese, Spanish, Chinese)
• Function calling support for tool use and agentic workflows