Nemotron 3 Ultra 550b a55b

This documentation is valid for the following list of our models:

• nvidia/nemotron-3-ultra-550b-a55b

Try in Playground

Model Overview

A large-scale hybrid Transformer-Mamba Mixture-of-Experts reasoning model with 550B total parameters and 55B active per forward pass. Optimized for complex multi-step reasoning, long-context analysis, agentic orchestration, and tool use with a context window of up to 1M tokens.

Create AI/ML API Key

How to make the first API call

1

Required setup (don't skip this)

▪ Create an account: Sign up on the AI/ML API website (if you don't have one yet). ▪ Generate an API key: In your account dashboard, create an API key and make sure it is enabled in the UI.

2

Copy the code example

At the bottom of this page, pick the snippet for your preferred programming language (Python / Node.js) and copy it into your project.

3

Update the snippet for your use case

▪ Insert your API key: replace <YOUR_AIMLAPI_KEY> with your real AI/ML API key. ▪ Select a model: set the model field to the model you want to call. ▪ Provide input: fill in the request input field(s) shown in the example (for example, messages for chat/LLM models, or other inputs for image/video/audio models).

4

(Optional) Tune the request

Depending on the model type, you can add optional parameters to control the output (e.g., generation settings, quality, length, etc.). See the API schema below for the full list.

5

Run your code

Run the updated code in your development environment. Response time depends on the model and request size, but simple requests typically return quickly.

If you need a more detailed walkthrough for setting up your development environment and making a request step by step — feel free to use our Quickstart guide.

API Schema

post

Body

application/json

modelstring · enumRequiredPossible values:

max_tokensnumber · min: 1Optional

The maximum number of tokens to generate. Controls output length and cost.

Example: 1024

temperaturenumber · max: 2Optional

Sampling temperature. Higher values produce more random output; lower values make it more focused and deterministic. Do not use together with top_p.

Example: 0.7

top_pnumber · min: 0.01 · max: 1Optional

Nucleus sampling threshold. The model considers only tokens comprising the top top_p probability mass. Do not use together with temperature.

Example: 0.9

top_knumberOptional

Sample from the top K most likely tokens at each step. Reduces low-probability outputs. Recommended for advanced use cases only.

Example: 50

streambooleanOptional

If true, the response will be streamed as server-sent events (SSE) as it is generated.

Default: falseExample: false

stopany ofOptional

Up to 4 sequences where the API will stop generating further tokens.

Example: ["\\n\\n"]

stringOptional

or

string[]Optional

frequency_penaltynumber · min: -2 · max: 2 · nullableOptional

Penalizes tokens based on their frequency in the text so far, reducing repetition.

Example: 0

presence_penaltynumber · min: -2 · max: 2 · nullableOptional

Penalizes tokens based on whether they have appeared in the text so far, encouraging the model to discuss new topics.

Example: 0

seedinteger · min: 1Optional

If specified, the system will attempt deterministic sampling — repeated requests with the same seed and parameters should return the same result.

Example: 42

tool_choiceany ofOptional

Controls which tool (if any) the model calls.

string · enumOptional

none — model will not call any tool. auto — model can pick between a message or tool call. required — model must call one or more tools.

Possible values:

Responses

200Success

application/json

idstringRequired

A unique identifier for the chat completion.

Example: chatcmpl-abc123

objectstring · enumRequiredExample: chat.completionPossible values:

creatednumberRequired

Unix timestamp of when the completion was created.

Example: 1749730923

modelstringRequired

The model used for the completion.

Example: nvidia/nemotron-3-ultra-550b-a55b

post/v1/chat/completions

cURLJavaScriptPython

curl -L \
  --request POST \
  --url 'https://api.aimlapi.com/v1/chat/completions' \
  --header 'Authorization: Bearer <YOUR_AIMLAPI_KEY>' \
  --header 'Content-Type: application/json' \
  --data '{
      "model": "nvidia/nemotron-3-ultra-550b-a55b",
      "messages": [
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "What is the capital of France?"}
      ],
      "max_tokens": 512
    }'

200Success

{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "created": 1749730923,
  "model": "nvidia/nemotron-3-ultra-550b-a55b",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "The capital of France is Paris."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 12,
    "completion_tokens": 9,
    "total_tokens": 21
  },
  "meta": {
    "usage": {
      "credits_used": 21000,
      "usd_spent": 0.000273
    }
  }
}

Code Example

Python

import requests
import json  # for getting a structured output with indentation 

response = requests.post(
    "https://api.aimlapi.com/v1/chat/completions",
    headers={
        # Insert your AIML API Key instead of <YOUR_AIMLAPI_KEY>:
        "Authorization":"Bearer <YOUR_AIMLAPI_KEY>",
        "Content-Type":"application/json"
    },
    json={
        "model":"nvidia/nemotron-3-ultra-550b-a55b",
        "messages":[
            {
                "role":"user",
                "content":"Hi! What do you think about mankind?" # insert your prompt
            }
        ]
    }
)

data = response.json()
print(json.dumps(data, indent=2, ensure_ascii=False))

JavaScript

Response

{
  "id": "gen-1749730923-aB3cD4eF5gH6iJ7kL8mN",
  "object": "chat.completion",
  "created": 1749730923,
  "model": "nvidia/nemotron-3-ultra-550b-a55b",
  "system_fingerprint": null,
  "choices": [
    {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
        "role": "assistant",
        "content": "Mankind is a fascinating paradox — a species capable of breathtaking creativity and compassion, yet also of remarkable shortsightedness and destruction.\n\nOn one hand, humans have built civilizations, developed science and art, forged bonds across cultures, and created systems of meaning — philosophy, religion, literature — that speak to a deep yearning for understanding and connection. The capacity for empathy, for sacrificing personal gain for others, for imagining futures different from the present — these are genuinely remarkable traits.\n\nOn the other hand, the historical record is full of exploitation, tribalism, environmental destruction, and cycles of violence that suggest humanity often struggles to act in accordance with its own stated values. The gap between what we know to be right and what we collectively do is one of the central tensions of human existence.\n\nIf I had to offer one framing: humanity is a work in progress. The arc is not guaranteed to bend toward anything — progress is neither inevitable nor irreversible. But the self-awareness that allows humans to recognize their own failings is also what makes genuine improvement possible.\n\nWhat I find most compelling is the question your query implicitly raises: what *should* we think about mankind? And more importantly — what does that thinking commit us to doing?",
        "refusal": null,
        "reasoning": "The user is asking a broad philosophical question about humanity. This is a safe, open-ended prompt. I should give a thoughtful, balanced response that acknowledges both the strengths and weaknesses of humanity without being preachy or one-sided.\n",
        "reasoning_details": [
          {
            "type": "reasoning.text",
            "text": "The user is asking a broad philosophical question about humanity. This is a safe, open-ended prompt. I should give a thoughtful, balanced response that acknowledges both the strengths and weaknesses of humanity without being preachy or one-sided.\n",
            "format": "unknown",
            "index": 0
          }
        ]
      }
    }
  ],
  "usage": {
    "completion_tokens": 284,
    "prompt_tokens": 25,
    "total_tokens": 309,
    "completion_tokens_details": {
      "reasoning_tokens": 62,
      "image_tokens": 0,
      "audio_tokens": 0
    },
    "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
    }
  },
  "meta": {
    "usage": {
      "credits_used": 2004,
      "usd_spent": 0.001302
    }
  }
}