Text Generation Model

Nemotron 3 Ultra 550B

Nemotron 3 Ultra 550B is a text generation model from Nvidia with a 262,144-token context window and reasoning support.

Start Building with Nemotron 3 Ultra 550B View All Models

Publisher

Nvidia

TypeText

Context Window262,144 tokens

ReleasedMay 2026

Input$0.50/MTok

Output$2.20/MTok

Provider

DeepInfra

FLAGSHIPREASONING

Try Nemotron 3 Ultra 550B →

About Nemotron 3 Ultra 550B

Large-scale reasoning with extended context

Nemotron 3 Ultra 550B is a 550-billion-parameter mixture-of-experts language model developed by Nvidia, with 55 billion parameters active per forward pass. It is designed for text generation tasks and supports a context window of 262,144 tokens, making it suitable for processing long documents, extended conversations, and complex multi-step tasks. The model is available through DeepInfra and includes a configurable reasoning mode that can be toggled as an input parameter.

The model is tagged as a flagship reasoning model, indicating it is positioned for tasks that benefit from structured, multi-step inference. With a maximum response size of 16,384 tokens and a large active parameter count, it is suited for applications such as document analysis, code generation, summarization of lengthy inputs, and complex question answering. The reasoning input toggle allows developers to control whether the model applies extended reasoning behavior, offering flexibility depending on the use case.

Capabilities

What Nemotron 3 Ultra 550B supports

Extended Context Window

Supports up to 262,144 tokens of context, enabling processing of long documents, codebases, or multi-turn conversations in a single pass.

Configurable Reasoning

Includes a selectable reasoning mode input that lets developers enable or disable extended multi-step reasoning behavior per request.

Large Response Output

Supports a maximum response size of 16,384 tokens, allowing for detailed, long-form completions in a single generation.

Mixture-of-Experts Architecture

Uses a sparse mixture-of-experts design with 550 billion total parameters and approximately 55 billion active per forward pass, balancing capacity with compute efficiency.

Text Generation

Generates natural language text for tasks including summarization, question answering, document analysis, and code generation.

Ready to build with Nemotron 3 Ultra 550B?

Get Started Free

FAQ

Common questions about Nemotron 3 Ultra 550B

What is the context window size for Nemotron 3 Ultra 550B?

Nemotron 3 Ultra 550B supports a context window of 262,144 tokens, allowing it to process very long inputs in a single request.

What does the reasoning input toggle do?

The model includes a selectable 'Reasoning' input parameter. When enabled, the model applies extended multi-step reasoning behavior. This can be toggled per request depending on the task requirements.

How many parameters does Nemotron 3 Ultra 550B have?

The model has 550 billion total parameters and uses a mixture-of-experts architecture with approximately 55 billion parameters active per forward pass, as indicated by the model ID suffix 'A55B'.

What is the maximum response length this model can generate?

The model supports a maximum response size of 16,384 tokens per generation.

Is pricing information available for Nemotron 3 Ultra 550B on MindStudio?

Published pricing details are not listed in the current metadata. You can check MindStudio or the DeepInfra provider page for up-to-date pricing information.

What is the knowledge cutoff date for this model?

A specific knowledge cutoff date is not included in the available metadata. The model was released in May 2026; consult Nvidia's official documentation for details on training data recency.

Resources