# Fireworks on NVIDIA GPUs | Next-Gen AI Performance

> Build and scale generative AI with Fireworks on NVIDIA Blackwell and H200 GPUs—delivering 3.5× faster inference, Day-0 NVIDIA model access, and unmatched reliability.

## Fireworks on NVIDIA Accelerated Infrastructure

Fireworks on NVIDIA GPUs empowers you to build groundbreaking AI experiences leveraging the industry’s fastest inference engine and the most advanced, reliable GPUs

## State-of-the-Art GPU Hardware

Fireworks runs on the latest NVIDIA GPU architectures, delivering unprecedented performance for generative AI workloads

- **NVIDIA DGX B200**: NVIDIA DGX B200 is the world’s first system with the NVIDIA Blackwell GPU, delivering breakthrough performance for the world’s most complex AI problems, such as large language models and natural language processing. Configured with eight NVIDIA Blackwell GPUs, DGX B200 delivers unparalleled generative AI performance with a massive 1.4 terabytes (TB) of GPU memory, 64 terabytes per second (TB/s) of HBM3e memory bandwidth, and 14.4 TB/s of all-to-all GPU bandwidth, making it uniquely suited to handle any enterprise AI workload. · [Learn more about DGX B200](https://www.nvidia.com/en-us/data-center/dgx-b200/)
- **NVIDIA H200 GPU**: Built on the NVIDIA Hopper architecture, H200 delivers breakthrough performance for large language models and generative AI. · [Learn more about H200](https://www.nvidia.com/en-us/data-center/h200/)

## NVIDIA Nemotron: Optimized Open-Source Intelligence on Fireworks

NVIDIA Nemotron is a family of open AI models, engineered for high intelligence, compute efficiency, and deployment flexibility across a wide range of enterprise and developer workloads.

### Day 0 Nvidia Model Launches

Run state-of-the-art NVIDIA Nemotron models across multiple modalities from Day 0. As an NVIDIA Inception partner, Fireworks provides immediate access to the latest NVIDIA models the moment they're released:

### Text Models

Deploy NVIDIA Nemotron language models, including the latest Nemotron 3 Super, a 120B hybrid MoE model optimized for multi-agent AI systems.

Fireworks provides fast, scalable inference so you can bring Nemotron-powered applications to production quickly.

### Vision-Language Models

Deploy multimodal AI applications using NVIDIA Nemotron vision-language models, designed for tasks such as intelligent document processing, AI assistants with visual understanding, video captioning, and multi-modal agentic workflows.

## NVIDIA Nemotron Model Collection

NVIDIA Nemotron provides the most efficient open models, [powered by hybrid Mamba‑Transformer MoE](https://developer.nvidia.com/blog/inside-nvidia-nemotron-3-techniques-tools-and-data-that-make-it-efficient-and-accurate/) with 1M-token context, delivering top accuracy for complex agents.

- **[NVIDIA Nemotron 3 Super 120B A12B FP8](https://fireworks.ai/models/fireworks/nvidia-nemotron-3-super-120b-a12b-fp8)**: NVIDIA · 262,144 Context · LLM
- **[NVIDIA Nemotron Nano 2 VL](https://fireworks.ai/models/fireworks/nemotron-nano-v2-12b-vl)**: NVIDIA · 131,072 Context · Vision
- **[NVIDIA Nemotron Nano 3 30B A3B](https://fireworks.ai/models/fireworks/nemotron-nano-3-30b-a3b)**: NVIDIA · LLM
- **[NVIDIA Nemotron Nano 12B v2](https://fireworks.ai/models/fireworks/nvidia-nemotron-nano-12b-v2)**: NVIDIA · 128,000 Context · LLM
- **[NVIDIA Nemotron Nano 9B v2](https://fireworks.ai/models/fireworks/nvidia-nemotron-nano-9b-v2)**: NVIDIA · 128,000 Context · LLM
- **[NVIDIA Nemotron 3 Super 120B A12B NVFP4](https://fireworks.ai/models/fireworks/nvidia-nemotron-3-super-120b-a12b-nvfp4)**: NVIDIA · 262,144 Context · LLM

## NVIDIA Nemoclaw Integration

Autonomous AI agents are here, but deploying them securely at scale is challenging. Fireworks is collaborating with NVIDIA on **NVIDIA NemoClaw**, an open-source stack that simplifies running always-on agents safely. NemoClaw installs the NVIDIA OpenShell runtime to enforce policy-based security and privacy. As a day-0 inference provider, Fireworks delivers the speed and efficiency you need to deploy your agents immediately.

[NVIDIA Nemoclaw on Fireworks](https://www.nvidia.com/nemoclaw/)

## NVIDIA NIM Integration

Deploy and run models seamlessly using NVIDIA NIM inference microservices with Fireworks. NIM provides optimized inference containers that simplify deployment while maximizing performance on NVIDIA GPUs.

[NVIDIA NIM on Fireworks](https://fireworks.ai/blog/nvidia-nim-on-fireworks)

## The Ultimate AI Platform

Fireworks and NVIDIA together deliver the ultimate platform for generative AI, empowering developers with.state-of-the-art GPU hardware and the industry's fastest inference engine. This combination delivers unmatched performance, reliability, and scalability. Whether you're building conversational AI, vision applications, or complex multimodal systems, Fireworks on NVIDIA GPUs provides the essential foundation for innovating at scale, and delivering exceptional AI experiences.