Nemotron 3 Ultra
Nemotron 3 Ultra is a large language model made by NVIDIA, added to the catalog on 4 Jun 2026. It has a context window of 262,144 tokens and costs $0.50 per 1M input tokens and $2.20 per 1M output tokens. It accepts text and produces text. Typical uses: analysing lengthy documents, working through multi-step reasoning prompts and tool-assisted text workflows.
Price checked ยท source: OpenRouter
What Nemotron 3 Ultra is
NVIDIA Nemotron 3 Ultra is a text-in, text-out model from NVIDIA, described as a reasoning and orchestration model. It accepts text and produces text, and its listed supported features include reasoning, tool use, structured outputs, and response formatting. The context window holds up to 262,144 tokens, and the maximum output is 16,384 tokens. Those limits are relevant when evaluating it for prompts with substantial text input or responses that need room for a longer explanation.
The model is described as a mixture-of-experts system with a hybrid Transformer-Mamba architecture. The catalog lists 550 billion total parameters and 55 billion active parameters. Its context capacity and supported interaction options suit text tasks involving extended inputs, multi-step reasoning, or coordinating responses with tools. Structured output and response-format support are relevant when an application needs results arranged in a specified format rather than only free-form prose.
The listed modality is text for both input and output; the catalog does not list image or audio support. Potential uses include reviewing long reports, answering questions based on extensive text, preparing structured responses, and tool-assisted workflows. The model's stated input, output, and context limits can help determine whether it fits a particular text-based process. Its description identifies reasoning and orchestration as its focus, while the available facts do not specify particular domain expertise or task-specific performance.
Key features
- Accepts text and produces text
- Supports a context of up to 262,144 tokens
- Maximum output length is 16,384 tokens
- Supports reasoning, tools, structured outputs, and response formatting
Typical uses
- Analysing lengthy documents
- Working through multi-step reasoning prompts
- Tool-assisted text workflows
- Generating structured text responses
Nemotron 3 Ultra pricing
US dollars, list price through OpenRouter.
| Item | Standard | Free |
|---|---|---|
| Input, per 1M tokens | $0.50 | $0 |
| Output, per 1M tokens | $2.20 | $0 |
| Cached input read, per 1M tokens | $0.10 | n/a |
| Variant | Model id |
|---|---|
| Free | nvidia/nemotron-3-ultra-550b-a55b:free |
A variant is the same model under another id, with prices or limits of its own. Its prices are the columns of the list above.
No price change since it was added on 4 Jun 2026.
| What a task costs | Cost |
|---|---|
| Answer a short question500 in / 300 out | $0.0009 |
| Summarise a 50-page PDF25,000 in / 800 out | $0.014 |
| Review a 2,000-line source file30,000 in / 2,000 out | $0.019 |
| Read a 300-page contract set150,000 in / 1,500 out | $0.078 |
Nemotron 3 Ultra specifications
| Model id | nvidia/nemotron-3-ultra-550b-a55b |
|---|---|
| Version | nemotron-3-ultra-550b-a55b-20260604 |
| Context window | 262,144 tokensAbout 196,608 English words |
| Max output | 16,384 tokens |
| Knowledge cutoff | Not published |
| Added to catalog | 4 Jun 2026 |
| Input types | text |
|---|---|
| Output types | text |
| Tool calling | Yes |
| Structured outputs | Yes |
| Reasoning effort control | Yes |
| Content moderation | Not applied by provider |
Sample outputs
The same prompts are given to every model, and the unedited output is shown with the date it was produced. Upend does not publish leaderboard scores it did not run.
No output has been recorded for Nemotron 3 Ultra yet.
Compare Nemotron 3 Ultra
Side by side with other models: prices, context window, what each accepts and supports, and the same four tasks priced on both.