Free 24-hour trial
Private
Prompts are not logged
Not used for training

Try gpt-oss-20b

Use gpt-oss-20b free for 24 hours. Start the trial in the chat, ask a short question, then use the same key in your own app.

The trial ends after 24 hours, or after 5,000,000 tokens, whichever comes first. About 90 tokens/sec. Prompts stay on our hardware. They are not logged and not used to train models.

gpt-oss-20b

Free for 24 hours after you start

Start your free 24-hour trial

Enter your email and start the trial. The key lasts 24 hours and is shown once.
Try a question. Send stays off until the trial key exists.

Tokens served

—

Requests served

—

Trials issued

—

Slots

Loading…

Two trials can run at once. An occupied slot shows when it frees. Emails and keys are not listed. Model in service: gpt-oss-20b.

Use the key

Base URL https://llm.chargebite.net/v1. Model gpt-oss-20b. If the form above shows a different base URL, use that one.

Chat

A chat window on your computer.

Open WebUI

Admin Panel → Settings → Connections.
Under OpenAI, set the URL to https://llm.chargebite.net/v1 and paste the key.
Select model gpt-oss-20b.

Cherry Studio

Settings → Model Provider → Add.
Choose OpenAI-compatible. Set the API address to https://llm.chargebite.net/v1 and paste the key.
Add model gpt-oss-20b.

LibreChat

In librechat.yaml:

endpoints:
  custom:
    - name: Charge Bite
      apiKey: sk-…
      baseURL: https://llm.chargebite.net/v1
      models:
        default: [gpt-oss-20b]
        fetch: false

Chatbox

Settings → Add model provider → OpenAI API Compatible.
API Host: https://llm.chargebite.net/v1. Paste the key.
Model: gpt-oss-20b.

Coding

An editor or a coding agent.

Cursor

Settings → Models → API keys.
Turn on OpenAI and paste the key.
Set Override OpenAI Base URL to https://llm.chargebite.net/v1.
Add model gpt-oss-20b and select it in chat.

OpenClaw

In ~/.openclaw/openclaw.json:

{
  "agents": { "defaults": { "model": { "primary": "chargebite/gpt-oss-20b" } } },
  "models": {
    "mode": "merge",
    "providers": {
      "chargebite": {
        "baseUrl": "https://llm.chargebite.net/v1",
        "apiKey": "sk-…",
        "api": "openai-completions",
        "models": [{ "id": "gpt-oss-20b", "name": "gpt-oss-20b" }]
      }
    }
  }
}

Hermes

In ~/.hermes/config.yaml:

model:
  default: gpt-oss-20b
  provider: custom
  base_url: https://llm.chargebite.net/v1
  api_key: sk-…

OpenCode

In ~/.config/opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "model": "chargebite/gpt-oss-20b",
  "provider": {
    "chargebite": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Charge Bite",
      "options": {
        "baseURL": "https://llm.chargebite.net/v1",
        "apiKey": "sk-…"
      },
      "models": { "gpt-oss-20b": { "name": "gpt-oss-20b" } }
    }
  }
}

Cline

API Provider: OpenAI Compatible.
Base URL: https://llm.chargebite.net/v1
API Key: your trial key. Model ID: gpt-oss-20b.

Aider

export OPENAI_API_BASE=https://llm.chargebite.net/v1
export OPENAI_API_KEY=sk-…
aider --model openai/gpt-oss-20b

Continue (VS Code)

Add this to config.yaml:

models:
  - name: Charge Bite
    provider: openai
    model: gpt-oss-20b
    apiBase: https://llm.chargebite.net/v1
    apiKey: sk-…

Your own app

Call the API from your own code.

OpenAI Python SDK

from openai import OpenAI

client = OpenAI(
    base_url="https://llm.chargebite.net/v1",
    api_key="sk-…",
)
client.chat.completions.create(
    model="gpt-oss-20b",
    messages=[{"role": "user", "content": "Reply with pong"}],
)

Check that it answers

curl https://llm.chargebite.net/v1/chat/completions \
  -H "Authorization: Bearer sk-…" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-oss-20b","max_tokens":64,"messages":[{"role":"user","content":"Reply with pong"}]}'

Limits

5,000,000 tokens total. Prompts and replies both count.
24 hours, or until the tokens run out.
20 requests per minute. One reply at a time.
Two keys can be active. A third signup waits until a slot frees.
gpt-oss-20b only. About 90 tokens/sec when the model is loaded. Short questions fit this machine.

FAQ

Can I try gpt-oss-20b on this page?

Yes. Enter your email, generate a key, then ask a short question in the chat. Send stays off until the key exists. The reply stays in this browser tab and is not saved on the server.

Which apps can use the key?

Chat apps: Open WebUI, Cherry Studio, LibreChat, and Chatbox. Coding tools: Cursor, OpenClaw, Hermes, OpenCode, Cline, Aider, and Continue. The same key works with the OpenAI Python SDK. Base URL https://llm.chargebite.net/v1 and model gpt-oss-20b.

What is the ChargeBite LLM API base URL?

https://llm.chargebite.net/v1. After you get a key, ask a short question on this page. The same key also works at /chat/completions with Authorization: Bearer and model gpt-oss-20b.

Which model can I call?

gpt-oss-20b only. Any other model name is rejected.

How long does a free trial last?

24 hours from the moment the key is issued, or until 5,000,000 tokens are used, whichever comes first. Input and output tokens both count. The key is shown once.

What are the token and rate limits?

Each trial key includes 5,000,000 tokens total, across every request. Ask a short question. This machine answers best inside a few thousand tokens. Requests are limited to 20 per minute and one in-flight generation per key. Send model gpt-oss-20b. Other model names are rejected.

How many trial slots are there?

Two. The page lists open slots and occupied slots, including when an occupied slot frees. It does not show who holds a slot.

Are prompts stored?

No. The prompt is sent to the model on our hardware for the reply. It is not logged, and it is not used to train models. File tools on chargebite.net are separate and still do not need an account.