FW-Nemotron-Lightning-3.5-30B-A3B

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-Lightning-3.5-30B-A3B is a Mixture-of-Experts language model from NVIDIA's Nemotron-H family, identified as a 30B-parameter model with 3B active parameters in the Fireworks description. It uses a hybrid Mamba-Transformer architecture with interleaved Mamba-2 and MoE layers and select attention layers for efficient long-context inference. The model generates a reasoning trace before its final response by default, with reasoning behavior configurable through the chat template, and includes a Multi-Token Prediction (MTP) speculative decoding head for low-latency serving. On Fireworks, the model supports text input and output, function calling, serverless and on-demand deployment, and a 262,144-token endpoint context window.

• Hybrid Mamba-2 + MoE architecture with select attention layers
• 30B model class with 3B active parameters
• Configurable reasoning mode for direct responses or reasoning traces
• Multi-Token Prediction speculative decoding head for low-latency serving
• Multilingual support (English, French, Spanish, Italian, German, Japanese)
• Function calling support for tool use and agentic workflows
• Text input and output through Fireworks with a 262,144-token endpoint context window