> ## Documentation Index
> Fetch the complete documentation index at: https://docs.kymaapi.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Nemotron 3 Ultra 550B

> NVIDIA's strongest US open-weight model — 550B MoE with a hybrid Mamba-Transformer design, 1M context, and high throughput.

## Overview

`nemotron-3-ultra-550b` is NVIDIA's flagship open-weight model and the strongest US open-weight on the Artificial Analysis index (AA Index 48). It is a 550B mixture-of-experts model (55B active parameters) with a hybrid Mamba-Transformer architecture, a 1M context window, and throughput above 300 tokens/second. Reach for it when you want top open-weight reasoning and coding with fast generation.

## Specs

| Field              | Value                                                          |
| ------------------ | -------------------------------------------------------------- |
| Model ID           | `nemotron-3-ultra-550b`                                        |
| Best for           | Reasoning, coding, general work, long context, fast throughput |
| Context window     | 1M                                                             |
| Max output tokens  | 32K                                                            |
| Input modalities   | Text                                                           |
| Output modalities  | Text                                                           |
| Tool calling       | Yes                                                            |
| Structured outputs | Yes                                                            |
| Reasoning          | Yes                                                            |
| Prompt caching     | Yes                                                            |
| Speed              | Fast                                                           |
| Cost band          | Balanced                                                       |
| Release stage      | Stable                                                         |

## Use this when

* You want the strongest US open-weight model for reasoning and coding.
* You need high throughput on long generations.
* You need a very large (1M) context window.
* You build agents that need reliable tool calling and JSON outputs.

## Pick something else when

* You want the strongest tool-heavy agent: use [`kimi-k2.6`](/models/kimi-k2.6).
* You want the highest general quality: use [`qwen-3.7-plus`](/models/qwen-3.7-plus).
* You want deep step-by-step reasoning at lower cost: use [`deepseek-r1`](/models/deepseek-r1).

## Example

```python theme={null}
from openai import OpenAI

client = OpenAI(base_url="https://kymaapi.com/v1", api_key="ky-...")

response = client.chat.completions.create(
    model="nemotron-3-ultra-550b",
    messages=[{"role": "user", "content": "Derive the time complexity and propose a faster algorithm."}]
)
```

## Agent query example

```bash theme={null}
curl "https://kymaapi.com/v1/models?recommended_for=reasoning&tools=true&quality_tier=frontier-open"
```