FW-Nemotron-3-Ultra-NVFP4

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-3-Ultra-550B-A55B-NVFP4 is a frontier-scale large language model (LLM) trained by NVIDIA, designed to deliver strong agentic, reasoning, and conversational capabilities. It is optimized for demanding workloads, including complex multi-step agents and reasoning over code, math, and science. The model employs a hybrid Latent Mixture-of-Experts (LatentMoE) architecture with interleaved Mamba-2 and MoE layers, select attention layers, and Multi-Token Prediction (MTP). It is trained using an NVFP4 pre-training recipe and has approximately 550B total parameters with 55B active parameters. On Fireworks, the model is available through the serverless API with text input and output, function calling, and a 262,144-token endpoint context window.

• Hybrid Mamba-2 + LatentMoE architecture with select attention and MTP layers
• Approximately 550B total parameters with 55B active parameters
• Configurable reasoning mode for agentic and conversational tasks
• Multilingual support (English, French, Spanish, Italian, German, Japanese, Hindi, Korean, Brazilian Portuguese, Chinese)
• Function calling support for tool use and agentic workflows
• Text input and output through the Fireworks serverless API with a 262,144-token endpoint context window