One gateway for frontier models, at published per-token rates.
Point the client library you already use at our gateway and pay for the tokens you spend. Prepaid balance, no subscription, no monthly minimum.
- Published rates per model and token type: what a standard account is charged
- One key; one base URL for every model, plus the Messages API for text models
- Pay as you go, from a prepaid balance
- 37models in the catalogue
- 30available today
- Per tokenbilling, from a prepaid balance
Product
What you get
Properties of the system, not promises.
Your own client
One key and one gateway. An OpenAI-compatible library pointed at /api/v1 reaches every model, text and image alike; a Messages API client pointed at the gateway root reaches the text models in their native shape. Nothing is reinstalled, and no account with the model vendor is needed.
Learn morePer-token billing
Input, output, cache write and cache read are counted separately, at the published rate of the model the request used. Image models are metered the same way: the frame is output tokens, and the count depends on the frame shape. A refused request costs nothing.
Learn moreSpend controls
A prepaid balance and as many keys as you need, each revocable on its own. On request, a daily and a monthly ceiling in dollars on the whole account.
Learn moreA log you can check
Every answered request with its model, its four token counters and its charge.
Bodies are not stored
Prompts and answers stream through. We keep the model, the counters, the charge and the time.
Ceilings on the account
A daily and a monthly ceiling in dollars can be set on the account on request, shared by all your keys. A catch for a leaked key.
Several keys, each revocable
One per project or machine. Revoke any of them in your account; it stops working at once.
One pool per model
Refusals are retried across our pool. When nothing is free you get a 503, never a different model.
No subscription
You pay for tokens spent, not for a month. An idle balance does not expire.
Cache priced as cache
Cache writes and cache reads have their own rates in the catalogue — and in an agent session they are most of the tokens.
A person answers
Your message goes straight to the people who run the gateway, not to a ticket robot.
Models
One gateway, many models
One key reaches all of them. Open any model to see its pricing, its route and where it stands.
- Fable 5.1
- Fable 5
- Opus 5.5
- Opus 5
- Opus 4.8
- Opus 4.7
- Opus 4.6
- Sonnet 5.5
- Sonnet 5
- Sonnet 4.6
- Haiku 5.5
- Haiku 4.5
- GPT-6 AstraOpenAI
- GPT-6.1 SolOpenAI
- GPT-6.1 TerraOpenAI
- GPT-6 SolOpenAI
- GPT-6 LunaOpenAI
- GPT-5.6 SolOpenAI
- GPT-5.6 TerraOpenAI
- GPT-5.6 LunaOpenAI
- GPT Image 2.5 FlareOpenAI
- GPT Image 2.5 SunburstOpenAI
- GPT-5.4 Image 2OpenAI
- GPT-5 ImageOpenAI
- GPT-5 Image MiniOpenAI
- Nano Banana ProGoogle
- Nano Banana 2Google
- Nano Banana 2.1Google
- Nano Banana 2 LiteGoogle
- Nano BananaGoogle
- Gemini 3.8 FlashGoogle
- Gemini 3.5 Flash LiteGoogle
- DeepSeek V4 ProDeepSeek
- Llama 4 MaverickMeta
- Mistral Large 4Mistral AI
- Grok 4.7xAI
- GPT-5.5OpenAI
Getting started
Setup in three steps
Nothing is reinstalled: the client library you already use, pointed at a different base URL.
- 1
Sign up and add funds
Create an account, top up the balance from the account page, and create a key — several, if you like. - 2
Point your client at the gateway
Set the base URL — /api/v1 for an OpenAI-compatible client, the gateway root for a Messages API client — and pass the key as a bearer token. Each library takes exactly these two settings. - 3
Send requests
Each one is billed per token at the model’s published rate and shows up in your log with its counters and its charge.
Get started
Prepaid balance, topped up from your account. Unused balance is refunded on request, within up to 10 business days — write to us from your account email.