NVIDIA Nemotron 3 Ultra 550B A55B

latest

Model IDnvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4

NVIDIA Nemotron-3-Ultra-550B-A55B is the flagship open-weight model in the Nemotron 3 family — a 550B hybrid LatentMoE (Mamba-2 + MoE + Attention) with 55B active parameters, Multi-Token Prediction (MTP), and native NVFP4 pretraining. Built for frontier reasoning, complex multi-agent workflows, and long-context analysis with a native 1M-token context window and configurable thinking mode.

1M Tokens ContextFast InferenceCodingFunction CallingMultilingual

Save 20%

$1.44$1.15

/ 1M input tokens

$4.31$3.45

/ 1M output tokens

Get API Key

Request (Non streaming)

Open in Playground

from openai import OpenAI

# Initialize the OpenAI client with Qubrid base URL
client = OpenAI(
    base_url="https://qubrid.com/v1",
    api_key="QUBRID_API_KEY",
)

response = client.chat.completions.create(
    model="nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4",
    messages=[
      {
        "role": "user",
        "content": "Explain the main benefits of using a chat completion API for text generation."
      }
    ],
    max_tokens=16000,
    temperature=1,
    top_p=0.95,
    stream=False,
    extra_body={
        "enable_thinking": True,
    }
)

print(response.choices[0].message.content)

Example response

A chat completion API provides a standard way to send conversational input and receive model-generated text in a single request.

Key benefits include:

• Interoperability: any client can use HTTP with JSON request and response bodies.
• Flexibility: system prompts, user messages, and parameters such as temperature and max tokens are easy to configure.
• Observability: responses typically include token usage fields for cost and performance tracking.

A typical integration sends a POST request with the model name and messages array, then reads the assistant message from the first choice in the response.

Request (Non streaming)

Open in Playground

from openai import OpenAI

# Initialize the OpenAI client with Qubrid base URL
client = OpenAI(
    base_url="https://qubrid.com/v1",
    api_key="QUBRID_API_KEY",
)

response = client.chat.completions.create(
    model="nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4",
    messages=[
      {
        "role": "user",
        "content": "Explain the main benefits of using a chat completion API for text generation."
      }
    ],
    max_tokens=16000,
    temperature=1,
    top_p=0.95,
    stream=False,
    extra_body={
        "enable_thinking": True,
    }
)

print(response.choices[0].message.content)

Example response

A chat completion API provides a standard way to send conversational input and receive model-generated text in a single request.

Key benefits include:

• Interoperability: any client can use HTTP with JSON request and response bodies.
• Flexibility: system prompts, user messages, and parameters such as temperature and max tokens are easy to configure.
• Observability: responses typically include token usage fields for cost and performance tracking.

A typical integration sends a POST request with the model name and messages array, then reads the assistant message from the first choice in the response.

Streaming supported • Function calling supported • See all examples in Playground Open in Playground for streaming, files, and all parameters.

Open in Playground

Loading model details...

Fetching the latest models and pricing from the API.

A chat completion API provides a standard way to send conversational input and receive model-generated text in a single request. Key benefits include: • Interoperability: any client can use HTTP with JSON request and response bodies. • Flexibility: system prompts, user messages, and parameters such as temperature and max tokens are easy to configure. • Observability: responses typically include token usage fields for cost and performance tracking. A typical integration sends a POST request with the model name and messages array, then reads the assistant message from the first choice in the response.