cutetailbud/FastAIToken

What it is

FastAIToken is a paid API aggregation service for developers. One account and one API key can call models from several providers, including GPT, Claude, Gemini, Grok, DeepSeek, GLM and Kimi, plus video models such as Seedance, MiniMax H3, Wan and Veo. The repository is primarily the service's price list and onboarding guide, not a standalone client or gateway implementation.

Requests go through the service's base URL, so the same OpenAI-compatible SDK pattern works after you set base_url and api_key. The README publishes model names, context tiers and per-token or per-second prices for each group; the group multiplier is applied to the listed price.

Quick start

1. Register at the FastAIToken website and top up; the README states that 1 CNY buys 1 USD of credit.
2. Create an API key in the console and choose a model group.
3. Copy the base URL shown in the console.
4. Call the API with an OpenAI-compatible client:

from openai import OpenAI

client = OpenAI(
    api_key="sk-xxxxxxxx",
    base_url="<base URL from the console>",
)

resp = client.chat.completions.create(
    model="gpt-5.5",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(resp.choices[0].message.content)

The repository also provides an examples directory and a machine-readable pricing.json. The exact base URL, available groups and live prices are only available from the service console or website; the README does not substitute for them.

Pricing and billing

Billing is listed price × group multiplier. Groups differ in price and stated line stability: OpenAI groups range from a 0.06x promotional group to 0.5x enterprise lines, Claude groups are 1.5x and 1.9x, and domestic models and video are 1x. DeepSeek has a weekday peak multiplier in the stated GMT+8 windows. Prices in the repository are marked as reference only and may change without notice.

Constraints

This is a third-party relay, so availability, latency, rate limits and model behavior depend on the upstream providers and on which group you choose. The repository does not document SLA terms beyond group labels, and it does not contain the relay's server implementation. A key is required, so using the examples requires registration and payment. The license field is NOASSERTION in the repository metadata, so check the LICENSE file before redistributing anything from the repo.

HotRepo's take

On the Python new-project list for 2026-11-12, the project recorded 107 stars and 4 stars gained over a roughly one-hour window, with no forks. That is a very short observation period and does not establish adoption or reliability. For this project the more relevant fact is structural: the repository publishes prices and an integration example, while the actual service lives on the website, so the GitHub presence functions as documentation and an entry point rather than a self-contained tool.