One endpoint, the whole catalogue, a single API key.
Cerbero sits in front of several frontier providers and serves them as if they were one. You do not change library or the way you call: you change the base URL and carry on working.
cliente.py
# the only thing that changes
from openai import OpenAI
client = OpenAI(
- base_url="https://api.openai.com/v1",
+ base_url="https://api.cerberusapp.cc/v1",
api_key="sk-cb-···",
)
r = client.chat.completions.create(
model="claude-sonnet-5",
messages=[{"role": "user", "content": "Hello"}],
)Available models
This list comes from the live catalogue of the service, not from a hand-written copy. Any of these names works as-is in the model field of your call.
Status and response time are measured by our own probe every few minutes with a real request. Response is how long a short request takes end to end, not generation speed: most of that time is start-up, so a slow response does not mean the model writes slowly.
| Model | Plans | Status | Response |
|---|---|---|---|
| claude-sonnet-5 | BasicProEnterprise | Operational | 5.9 s |
| claude-sonnet-5-uncensored | BasicProEnterprise | Operational | 4.2 s |
| gemini-3.8-flash | BasicProEnterprise | Operational | 2.9 s |
| gemini-3.7-flash | BasicProEnterprise | Operational | 2.9 s |
| gemini-3.6-flash | BasicProEnterprise | Operational | 3.1 s |
| gemini-imagen | BasicProEnterpriseFree | Operational | 3.5 s |
| gemini-3.6-flash-uncensored | BasicProEnterprise | Operational | 4.3 s |
| claude-opus-4-6-Free | Free | Operational | 51.3 s |
Plans
The quota counts requests, not tokens, on a rolling 24-hour window. It does not reset at midnight: each request frees its own slot 24 hours after it was made.
Pro
Most popular1,500requests a day
- 7 models included
- 4 concurrent requests
- 60 USD a month
Enterprise
2,000requests a day
- 7 models included
- 7 concurrent requests
- 89 USD a month
About payment: the website does not process payments. Upgrades are handled on Telegram, where billing is already set up.