Every modelon one key,metered bythe token.

One OpenAI-compatible endpoint for Claude, GPT, Gemini and DeepSeek. Keep the SDK you already use, top up a prepaid balance, and pay only for the tokens you spend.

Input from $0.26 per million tokens, on gemini-3.1-flash-lite.

Get an API key Read the quickstart

No subscription and no monthly minimum. Top up from $5.

gemini-3.1-flash-lite

Settled

Sale

864 max out

Tokens

Per 1M, in / out

$0.26 / $1.58

Example requests at today’s prices. Pick a grade and hold the trigger to pour your own, or type “refill” to fill a tank.

Today’s prices

Each price is the provider’s rate plus a small markup, in USD per million tokens. Input and output are priced separately. Grades sort the catalog by output price, the way a pump sorts fuel.

Model

In / out per 1M tokens

In per 1M

Out per 1M

gemini-3.1-flash-liteStandard
gemini-3.1-flash-lite
Input $0.26, output $1.58 per million tokens.
gemini-3.5-flash-liteStandard
gemini-3.5-flash-lite
Input $0.32, output $2.63 per million tokens.
gpt-5.6-terraPremium
gpt-5.6-terra
Input $2.10, output $12.60 per million tokens.

See all 3 models and their prices Get an API key

What would your last month cost here?

Start from an example workload or enter the tokens you use in a day. The total uses the model’s listed price, so it is what your balance would pay.

gemini-3.1-flash-lite

Standard

30 days on Refillhub

$51.98

Per day: $1.73
Tokens in 30 days: 108M
Input per 1M: $0.26
Output per 1M: $1.58

Get an API key

One line changes

Keep the OpenAI SDK you already use. Point the base URL at Refillhub and swap the key; requests, streaming and tool calls keep the same shape.

Keys start with rh_ and live in your dashboard.

client.ts

import OpenAI from "openai"; const client = new OpenAI({  baseURL: "https://refillhub.biz.id/api/v1",  apiKey: process.env.REFILLHUB_API_KEY,}); const reply = await client.chat.completions.create({  model: "gemini-3.1-flash-lite",  messages: [{ role: "user", content: "Explain token metering." }],}); console.log(reply.choices[0].message.content);

How the balance works

Top up by card

Pick an amount from $5 and pay through Stripe. The credit shows on your balance immediately and does not expire.

Each request deducts its exact cost

Tokens are counted from the provider’s usage report and charged at the listed price, rounded up to the micro-dollar. Nothing runs past your balance.

You set the ceilings

Give each key a spend cap and a requests-per-minute limit. Every charge lands in Usage with the model, token counts, cost and latency.

Balance, USD

The balance drains with each request. Top up when it runs low.

Card top-upCredit added to balance+$10.00
gemini-3.1-flash-lite3,420 in, 1,512 out−$0.00328
gemini-3.5-flash-lite1,180 in, 864 out−$0.00264
gpt-5.6-terra640 in, 408 out−$0.006485

Example statement. Amounts are priced from today’s catalog rates.

Questions people ask first

Is the API OpenAI-compatible?: Yes. POST /api/v1/chat/completions takes the same body as OpenAI, including streaming and tool calls, and GET /api/v1/models lists the catalog.
What happens when my balance runs out?: The request returns 402 with code insufficient_balance and the dashboard points you at a top-up. If your balance cannot cover the requested output, max_tokens is capped to what it does cover rather than failing the call.
Which models can I call?: 3 models are live in the catalog right now, from Claude and GPT to Gemini, DeepSeek and open-weight models. Each one has its price per 1M tokens on the models page.
How is the price set?: Each model’s price is the provider’s rate plus a small fixed markup, published per 1M input and output tokens. No per-request fee and no minimum charge.
Do you store my prompts?: No. We record the model, token counts, cost, latency and status of each request so you can audit spend. Prompt and completion text is not stored.
Is there a free tier?: There is no free trial. Top up from $5 to make your first request.