About Tokenator
No marketing promises: what the service actually does, what it is made of, and which numbers you can verify on the site itself.
What Tokenator does
Tokenator is a gateway to language models from different vendors through a single API. You buy a token bundle, get a key and call models over an OpenAI- or Anthropic-compatible protocol. No separate account with each model vendor is required.
Internally the gateway does three things: it selects an upstream provider for the requested model, converts the protocol between formats when needed, and counts the tokens spent.
Service facts
| Models in the catalog | 41 |
| Model vendors | 14 |
| API formats | OpenAI-compatible and Anthropic-compatible |
| Who builds the models | anthropic, bytedance-seed, cursor, deepseek, gemini, meituan, minimax, moonshotai, openai, qwen, tencent, x-ai, xiaomi, z-ai |
The table is generated from the live service configuration and changes together with the catalog.
How it works
- Register in your account and buy a token bundle.
- The key appears in your account once payment is confirmed.
- Put the key and base URL into your tool or SDK — see integrations.
- Usage and the request log are visible in your account.
Routing and failover
A single catalog model can be backed by several upstream providers. A request goes to the highest-priority one; on a transport error, a 5xx, or an overload signal Tokenator tries the next. The client never sees this — it gets one response.
Current model state is on the status page.
What Tokenator does not do
- It does not train models or influence their quality — it is an access layer.
- It does not provide free access: usage is always charged against the key limit.
- It does not pass upstream provider error texts through to the client.