Nvidia

Use NVIDIA models with Agno agents.

NVIDIA's NIM API catalog serves models from NVIDIA and partners like Meta and Mistral, optimized for GPU-accelerated inference.

Explore NVIDIA’s models here.

Authentication

Set your NVIDIA_API_KEY environment variable. Get your key from Nvidia here.

export NVIDIA_API_KEY=***

These examples select Mistral Nemotron, which NVIDIA lists with a hosted API endpoint and function calling. The SDK default remains meta/llama-3.3-70b-instruct, whose free hosted endpoint is deprecated; partner and downloadable deployments are separate. Choose a model available to your NVIDIA account.

Example

Install the openai package:

uv pip install -U agno openai

Use Nvidia with your Agent:

from agno.agent import Agent
from agno.models.nvidia import Nvidia

agent = Agent(model=Nvidia(id="mistralai/mistral-nemotron"), markdown=True)

# Print the response in the terminal
agent.print_response("Share a 2 sentence horror story")
View more examples here.

Parameters

ParameterTypeDefaultDescription
idstr"meta/llama-3.3-70b-instruct"The id of the NVIDIA model to use
namestr"Nvidia"The name of the model
providerstr"Nvidia"The provider of the model
api_keyOptional[str]NoneThe API key for NVIDIA (defaults to NVIDIA_API_KEY env var)
base_urlstr"https://integrate.api.nvidia.com/v1"The base URL for the NVIDIA API

Nvidia extends the OpenAI-compatible interface and supports most parameters from the OpenAI model.