
NVIDIA Nemotron 3 Ultra
NVIDIA Nemotron 3 Ultra is NVIDIA's frontier-scale open model for reasoning, agentic workflows and long-context analysis. It's a 550B-parameter hybrid Mamba-Transformer mixture of experts with 55B active per token and up to 1M tokens of context.
Ways to run NVIDIA Nemotron 3 Ultra
Dedicated endpoint
Get a quote →Your own GPUs, tuned for your traffic. We profile your workload, tune the deployment and put throughput and latency in your contract.
Serverless
API docs →Pay per token on shared GPUs, with no capacity to plan or reserve.
Inside your network
Security →Run a dedicated endpoint inside your own network, including air-gapped environments.
Pricing
Use it with the agents we set up for you
Point any open-source agent at your endpoint, or pick from our catalog and we set it up as part of your contract.
NVIDIA Nemotron 3 Ultra FAQ
How much does NVIDIA Nemotron 3 Ultra cost on Saturn Cloud?
Serverless costs $0.15 per 1M input tokens and $0.60 per 1M output tokens. A dedicated endpoint is priced per GPU-hour for the GPUs it runs on, and we size the GPUs to your traffic.
What is Nemotron 3 Ultra's context length?
Up to 1,048,576 tokens. NVIDIA’s serving examples default to 262,144.
Can I turn off reasoning in Nemotron 3 Ultra?
Yes. Reasoning is on by default. Turn it off with enable_thinking in the chat template, and cap the reasoning trace with reasoning_budget.
Does Nemotron 3 Ultra support tool calling?
Yes. On vLLM and SGLang it uses the qwen3_coder tool-call parser.
What GPUs does Nemotron 3 Ultra need?
NVIDIA’s minimum is 8x GB200, B200, GB300 or B300, 16x H100, or 8x H200. On serverless it runs on our GPUs. On a dedicated endpoint we size the GPUs to your traffic.