Fireworks

Fireworks provides hosted, OpenAI-compatible endpoints for open large language models.

Authentication

Create a Fireworks API key, then set the FIREWORKS_API_KEY environment variable.

export FIREWORKS_API_KEY=***

setx FIREWORKS_API_KEY ***

Prompt caching

Prompt caching happens automatically when you use the Fireworks model. Read more about how Fireworks handles caching in their docs.

Example

Install the openai package:

uv pip install -U openai agno

Use Fireworks with your Agent:

from agno.agent import Agent
from agno.models.fireworks import Fireworks

agent = Agent(
    model=Fireworks(id="accounts/fireworks/models/gpt-oss-120b"),
    markdown=True
)

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

See Fireworks usage examples.

Parameters

Parameter	Type	Default	Description
`id`	`str`	`"accounts/fireworks/models/gpt-oss-120b"`	The ID of the Fireworks model to use
`name`	`str`	`"Fireworks"`	The name of the model
`provider`	`str`	`"Fireworks"`	The provider of the model
`api_key`	`Optional[str]`	`None`	The API key for Fireworks (defaults to FIREWORKS_API_KEY env var)
`base_url`	`str`	`"https://api.fireworks.ai/inference/v1"`	The base URL for the Fireworks API

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

Get Started

Basics

Advanced

Production

Providers

Other

Additional Resources

Authentication

Prompt caching

Example

Parameters

​Authentication

​Prompt caching

​Example

​Parameters

Authentication

Prompt caching

Example

Parameters