How to top up an API relay: billing, RMB payments and cost control (2026)

The recommendation: Legitimate API relays support RMB top-ups through WeChat Pay and Alipay, with token or monthly billing. Three steps fund the account; balances do not expire and every charge is visible. This guide covers billing, top-up precautions, cost control and invoices.

Introduction

Last month, independent AI-tool developer Zhou topped up 500 yuan on an "ultra-cheap" platform. It worked for three days, began returning errors on day four, and became inaccessible on day five. The 500 yuan became a costly lesson.

Topping up seems simple, but the wrong platform can cost you both money and service.Where you top up matters more than how much. This guide covers billing methods, top-up steps, records, cost control and invoices for choosing an API relay in 2026.

Key takeaways

  • Two billing models: pay as you go without monthly fees, or monthly subscriptions for predictable costs and quotas.
  • RMB top-ups through WeChat Pay and Alipay, with separate input/output token rates.
  • Balances do not expire; itemized console records can be filtered by key and date.
  • Be wary of "lifetime purchase" and "unlimited ultra-cheap usage" claims, which are common among platforms that disappear.
  • Five cost controls: key isolation and quotas, suitable model tiers, caching and batches, anomaly monitoring, and small initial tests.

Pay as you go or monthly billing?

Legitimate gateways such as Yomi API, covering 200+ models and 30+ providers, offer pay-as-you-go and monthly billing.

Pay as you go has no monthly fee and suits early-stage projects or variable workloads. Input and output tokens are metered separately, with model-specific rates. See the live Model Explorer pricing table. For model-by-model comparison, see this price-per-million-tokens table.

Monthly billing fixes costs and provides a stable quota for predictable, frequent usage. The two approaches can be combined: subscriptions for routine traffic and pay as you go for spikes.

DimensionPay as you goMonthly billing
Payment method No monthly fee; charges based on token usage Prepaid monthly fee
Best suited for Initial testing and variable usage Stable, frequent, budget-sensitive usage
Cost profile Pay only for what you use Capped cost and steady quota
Suggested approach Start small, validate, then scale Move to monthly billing after validation to reduce costs

Top up in three steps — and avoid common mistakes

Funding an account is straightforward:

  1. Visit the Yomi API website, register with email and log in to the console.
  2. Open the top-up page and select WeChat Pay or Alipay.
  3. Scan the payment code; start calling once the balance arrives.

Four precautions:

Treat "lifetime" and "unlimited ultra-cheap" offers with caution; disappearing platforms often use these claims. Putting a large balance into an unidentified platform hands over both your wallet and access to your data.

Regulation is tightening too. On June 8, 2026, the Ministry of State Security warned of exposed data, downgraded models, malicious implants and outbound data transfers. The Cyberspace Administration is also pursuing its AI application campaign. Shanghai operator "Guapi" was detained for 37 days in May 2026 over noncompliant operations, then released on bail.

The lesson is to investigate a platform before paying. For a full safety checklist, read Are API relays safe?.

Usage records: every charge accounted for

Legitimate platforms record where your money goes. The console usage page shows the model, token count, cost and time of every request.

Every charge can be found by key and date in the console. Filter by key for project spending and by date for reconciliation and budgets. TLS 1.3 protects transmission, request content is not stored and intelligent routing switches routes during failures, combining low-latency global access with clear accounting.

Five ways to control costs

Different usage patterns can multiply the cost of the same AI feature. Combine these five methods.

Separating project keys and setting quotas is one of the most effective cost controls.

  1. Match the model tier to the task: use lightweight models such as deepseek-v4-flash for simple tasks and flagships for complex reasoning; prices can differ severalfold.
  2. Caching and batches: cache repeated content and check available batch pricing in the console.
  3. Monitor anomalies: configure spending alerts, review records daily and immediately disable keys with abnormal usage.
  4. Validate before scaling: start with a small amount and confirm latency and stability before choosing the main model and budget.

Invoices and company identity: a selection baseline

Business developers should confirm invoice availability in advance.Invoice availability is a quick test of whether a platform is suitable for a lasting business relationship.

Legitimate platforms have an identifiable company and invoices matching amounts, names and itemized records, which also provide accountability in disputes. Avoid topping up where the operator or invoicing process is unclear.

Individual developers should also retain every top-up and usage record for their own protection.

Frequently asked questions (FAQ)

Four common questions:

Final thoughts

Returning to Zhou's 500-yuan loss, the real question was where to pay, not whether to pay.

Costs are controllable when the platform has an identifiable operator, clear records and customer support. Validate your first call with a small top-up. Keep the balance and charges visible, and switch models by changing one field if necessary.

The Model Explorer pricing table is always available for model comparisons, selection and integration.

Topping up need not be a gamble. Open the console, start small, run your first call and check the records. Stop if the service is unsuitable.

Open the console and start with a small top-up →