What should you check first in this article?

Model setup recipes

Try NVIDIA Nemotron's Free API: Send a First Request to the Developer Endpoint

Send a short Python request to the free development endpoint and inspect the response.

Choose a hosted model in NVIDIA API¹ Catalog, then send a question from Python with its model ID and an API key to receive an answer.

Send a text request and the hosted model returns an answer

NVIDIA API Catalog is a developer service for trying models hosted on NVIDIA servers through an API. Browse models and examples in the catalog, then send a question from Python using an API key and model ID. Here, we'll ask Nemotron 3.5 Lightning a short question and inspect its answer and finish_reason.

A compact computer sits beside a laptop chat and code on a monitor — The Python example asks for two points on what a 429 error means and what to try next, then shows the model’s answer in the terminal.

Check the free endpoint and model terms

The current catalog page for nvidia/nemotron-3.5-lightning-30b-a3b lists both Free Endpoint and Download Available. Use nvidia/nemotron-3.5-lightning-30b-a3b as the hosted API model ID; the public weights are also available on Hugging Face. The API Trial Terms govern use of the hosted endpoint, while OpenMDW License 1.1 applies to the model materials.

Nemotron 3.5 Lightning is a text model with 3B active parameters out of 30B total. Its model card explicitly lists English and Japanese among its supported languages, but not Korean. If your task requires Korean output, first check quality with a short, non-sensitive prompt².

Prepare the API key and Python environment

Create a key with Generate API Key on the catalog’s Build screen and store it in a password manager. This API processes input on NVIDIA servers. NVIDIA’s API Trial Terms say inputs and outputs may be used to provide and improve the service, and restrict confidential, sensitive, personal, health, and payment information. Use only a short, non-sensitive prompt in this example.

In a macOS or Linux shell, create a Python virtual environment³ and install the OpenAI-compatible SDK. In the next step, save quickstart.py and run it from the same terminal.

Prepare a virtual environment and the OpenAI SDK

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip openai

Install the SDK in this environment, then run the Python file below.

A robot uses a laptop while keys for documents and network access hang in an open cabinet — Before sending a request, check which service and permissions the API key belongs to.

Send a short request and read its completion status

quickstart.py reads the key at runtime⁴ without echoing it. It sets the catalog endpoint and model ID, then asks the model for two points explaining what a 429 error means and what to try next. The code prints the model's answer and finish_reason. max_tokens caps generation tokens⁵, while reasoning_budget limits the tokens allocated to reasoning.

Send a Python request to the Nemotron endpoint

# quickstart.py
from getpass import getpass
from openai import OpenAI

client = OpenAI(
    base_url="https://integrate.api.nvidia.com/v1",
    api_key=getpass("NVIDIA API key: "),
)

response = client.chat.completions.create(
    model="nvidia/nemotron-3.5-lightning-30b-a3b",
    messages=[{"role": "user", "content": "In two short bullets, explain what a 429 API error means and what to try next."}],
    max_tokens=8192,
    extra_body={"chat_template_kwargs": {"enable_thinking": True}, "reasoning_budget": 4096},
)

choice = response.choices[0]
if choice.message.content:
    print(choice.message.content)
else:
    print(f"No final text returned; finish_reason={choice.finish_reason}")
print(f"finish_reason={choice.finish_reason}")

Run `python quickstart.py` to print the model’s explanation and `finish_reason` in the terminal.

Run the example

python quickstart.py

Check the response text and completion status together.

Check key errors and the permitted use

On success, choice.message.content contains the model’s answer and finish_reason shows how generation ended. For 401, recheck the key in the catalog. For 429, wait and retry with more space between requests. Check the model page for current Free Endpoint status and account availability.

The Developer Program API endpoint is for development, testing, and evaluation. Public documentation does not specify monthly credits or a guaranteed request count; a service for real end users needs a separate production route.

A desk monitor displays a conversation panel beside a code editor. — If no answer text appears in the terminal, check `finish_reason` first.

Terminology notes

API — A defined interface that lets other code call a program’s functions.
Back to the text
Prompt — The input text sent to a model. For generative models it describes the desired result and strongly shapes the output.
Back to the text
Python virtual environment — An isolated space for installing Python packages per project, helping reduce package-version conflicts.
Back to the text
Runtime — The software environment that provides facilities needed while a program runs. In local AI, it can also refer to a model execution engine.
Back to the text
Token — A unit into which a model divides input or output for processing.
Back to the text

Frequently asked questions

What should you check first in this article?: Choose a hosted model in NVIDIA API Catalog, then send a question from Python with its model ID and an API key to receive an answer.