Documentation: API endpoints and usage

Overview

• This collection exposes multiple base URLs (providers) for chat-style completions. Each entry in the table lists a base URL (with /models removed) and whether an API key is required.
• Base URL extraction: remove /models from the URL in your table to get the API base_url you should use in requests.

Base URLs Table

Also Supported API Routes:

• Nvidia: https://integrate.api.nvidia.com/v1
• DeepInfra: https://api.deepinfra.com/v1
• OpenRouter: https://openrouter.ai/api/v1
• Google Gemini: https://generativelanguage.googleapis.com/v1beta/openai
• xAI: https://api.x.ai/v1
• Together: https://api.together.xyz/v1
• OpenAI: https://api.openai.com/v1
• Strinable Inf: https://stringableinf.com/api/v1
• TypeGPT: https://typegpt.ai/api
• ApiAirforce: baseUrl: https://api.airforce/v1, apiEndpoint: https://stringableinf.com/api/v1/chat/completions
• Auto Provider & Model Selection: apiEndpoint: https://g4f.dev/ai/{now}

Individual clients available for:

• Pollinations AI
• Puter AI
• HuggingFace
• Ollama

How to choose a base URL

• If you want a local or self-hosted instance, you can use:
  • https://localhost:1337/v1
• If you want a free or public provider, you can use one of the g4f.dev endpoints (e.g., groq, ollama, pollinations.ai, nvidia, grok).
• If you want Azure-backed usage or hosted instances, use:
  • https://g4f.dev/api/Azure (requires API key)
  • https://g4f.dev/v1 (requires API key)

API usage basics

• Endpoints: All chat-style interactions use the chat completions endpoint at {base_url}/chat/completions
• Authentication:
  • If the table entry says "none required", you do not need to pass an API key.
  • If an API key is required (Azure, hosted instances, or gpt-oss-120b), supply api_key in your client configuration or request headers.
• Payload shape (typical):
  • model: string (e.g., gpt-oss-120b, gpt-4o, etc.)
  • temperature: number (optional)
  • messages: array of { role: "system" | "user" | "assistant", content: string }

Example payload

{
  "model": "gpt-4o",
  "temperature": 0.9,
  "messages": [{"role": "user", "content": "Hello, how are you?"}]
}

Examples

1) Python requests (local example)

• Sends a chat completion to the local endpoint at /v1/chat/completions

import requests

payload = {
    "model": "gpt-4o",
    "temperature": 0.9,
    "messages": [{"role": "user", "content": "Hello, how are you?"}]
}

response = requests.post("http://localhost:1337/v1/chat/completions", json=payload)

if response.status_code == 200:
    print(response.text)
else:
    print(f"Request failed with status code {response.status_code}")
    print("Response:", response.text)

2) Python with OpenAI client (custom base_url)

• Use the OpenAI Python client but point it at your chosen base URL

from openai import OpenAI

client = OpenAI(
    api_key="secret",  #  A API key is required; set 'secrect' for "none required api_key" providers
    base_url="https://g4f.dev/api/gpt-oss-120b"  # replace with the chosen base_url
)

response = client.chat.completions.create(
    model="gpt-oss-120b",
    messages=[{"role": "user", "content": "Explain quantum computing"}],
)

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

3) JavaScript (HTML/JS client)

• Basic usage in a browser-like environment

<script type="module">
    import Client from 'https://g4f.dev/dist/js/client.js';

    // Initialize a client with a base URL and optional API key
    const client = new Client({ baseUrl: 'https://g4f.dev/api/grok', apiKey: 'secret' });

    const result = await client.chat.completions.create({
        model: 'grok-4-fast-non-reasoning',
        messages: [
            { role: 'system', content: 'You are a helpful assistant.' },
            { role: 'user', content: 'Tell me a joke.' }
        ]
    });
</script>

Notes and quick tips

• API Key Changes: The gpt-oss-120b endpoint now requires an API key. Retrieve it from g4f.dev/api_key.html.
• Hosted Instance URL: Updated from host.g4f.dev/v1 to g4f.dev/v1
• The examples assume a chat-style completions API where you pass messages and receive a response containing the assistant's content.
• The base_url is always the URL without the trailing /models segment