How the RelayRouter prepaid balance works: top-ups, deductions and running out

You add funds to RelayRouter by Stripe card, with a $0 platform fee, no minimum spend and no subscription. Each request is then deducted at the settlement rate of its model group. For example, the GPT group costs CNY 0.6 per $1 of standard usage and the Claude group costs CNY 2.0. Failed or errored requests are generally not billed. The sources for this article do not say what happens at a zero balance, so check your dashboard.

How do top-ups work?

You top up with a Stripe card payment, and no subscription or minimum spend is required. According to relayrouter.io, the platform charges a $0 platform fee, so the amount you add goes toward model usage. The sources for this article do not list fixed top-up amounts or currencies, so check the payment screen for the options available to you. After you add funds, create an API key at https://relayrouter.io/dashboard and send it with every request as Authorization: Bearer YOUR_API_KEY. Usage is charged to the account that owns the key, so keep keys private and create separate keys for separate projects if you want clearer tracking. There is no recurring fee, so if you make no requests, nothing is deducted.

How are deductions calculated?

Each request is deducted according to the pricing rule for its model group or model, as listed at relayrouter.io/models. Mainstream groups use a settlement rate per $1 of standard usage. That rate is compared against a market reference of CNY 6.8 per $1. Some models are instead priced directly per token. On average, mainstream model groups are about 30 percent below official list prices.

Which requests draw down the balance?

Successful requests sent with your key through any of the three supported protocols draw down the balance. Failed or errored requests are generally not billed. According to relayrouter.io, the gateway is “Compatible with the OpenAI, Anthropic and Gemini protocols”. The three endpoints are POST /v1/chat/completions (base https://relayrouter.io/v1), POST /v1/messages (base https://relayrouter.io) and POST /v1beta/models/{model}:generateContent. The catalog has about 108 models across 19 public groups. These include claude-opus-5-5, claude-fable-5-1, gpt-6-astra, gpt-5.6-sol and gemini-3.8-flash, plus DeepSeek, GLM, MiniMax and Moonshot models. Streaming is supported. Because rates are set per group or per model, the model you choose determines how quickly your balance goes down.

What happens when the balance runs out?

The sources for this article do not describe what RelayRouter does when a balance reaches zero. They do not specify whether requests are rejected, what error is returned or whether automatic top-ups exist. Check the relayrouter.io/docs or your dashboard to confirm. A practical routine is:

1. Check the live per-model rates at relayrouter.io/models before heavy usage.
2. Schedule deepseek-v4-flash jobs outside the weekday 09:00 to 12:00 and 14:00 to 18:00 Beijing time windows, when its prices double.
3. Monitor your balance in the dashboard and top up by Stripe card before it runs low.
4. Handle failed responses in your code, since failed requests are generally not billed but still need a retry.

FAQ

These three questions cover the points readers ask about most often.

Do I have to change my code to start using the balance?
No. According to the relayrouter.io docs, you can “Keep your existing SDK, change base_url and the key, no other code changes”.

Is there a monthly fee or minimum spend?
No. The platform fee is $0, and there is no minimum spend and no subscription.

Am I charged when a request errors?
Generally no. Failed or errored requests are generally not billed. Check the live rates and current billing terms at relayrouter.io/models.