Flat-rate LLM API One key. Flat price. No meter.

An OpenAI- and Anthropic-compatible API on open-weight models we host. Buy a time pass and send as many tokens and requests as you like until it ends.

Join the waitlist See endpoints

0token caps

0request caps

2parallel sessions

3API formats

Endpoints

OpenAI-compatible

Base URL: https://api.endlesstokens.ai/v1
POST: /chat/completions · /responses · /completions · /embeddings
GET: /models · /models/<name>

Anthropic-compatible

Base URL: https://api.endlesstokens.ai
POST: /v1/messages · /v1/messages/count_tokens
Clients: Claude Code, the Anthropic SDK and any Messages client

Send your key as Authorization: Bearer, x-api-key or api-key. Model names are case-sensitive: use them exactly as GET /v1/models lists them.

Try it

copy

OPENAI_BASE_URL=https://api.endlesstokens.ai/v1
OPENAI_API_KEY=<your-key>ANTHROPIC_BASE_URL=https://api.endlesstokens.ai
ANTHROPIC_AUTH_TOKEN=<your-key>curl https://api.endlesstokens.ai/v1/chat/completions \
  -H "Authorization: Bearer <your-key>" \
  -H "Content-Type: application/json" \
  -d '{"model": "<model-name>",\
       "messages": [{"role": "user", "content": "hello"}]}'

Chat Completions and Responses clients, SDKs and editors. Any client that lets you set a base URL and key works.

Limits, stated up front

TokensNot counted

RequestsNo cap

Parallel sessions2 per key

Context window128K

ImagesUp to 5 per request

Reasoning effortUp to medium

Metered, capped or time pass?

How a flat pass compares with per-token APIs and “unlimited” subscriptions: see the comparison.

Errors you may see

`401`

The key is wrong or revoked.

`402`

No active pass on the account.

`429`

More than two parallel sessions on the key.

`503 / 529`

Every GPU is busy. Retry after the Retry-After header; the time is credited.

FAQ

Which API formats do you support?

OpenAI Chat Completions, Responses, Completions and Embeddings, and Anthropic Messages, all with the same key.

Which header carries the key?

Any one of Authorization: Bearer, x-api-key or api-key works on every endpoint.

Which models?

Open-weight models we host ourselves. GET /v1/models lists the exact names and each context length.

What happens when every GPU is busy?

You get 503 in the OpenAI format or 529 in the Anthropic format, with a Retry-After header. Honour it; the rejected time is credited back.

Is there a subscription?

No. You buy a 1-day or 3-day pass. It does not auto-renew.

One key. Endless tokens.

Join the waitlist See pricing