cheapaimodelSearch
All posts

DeepSeek API Payment: PayPal, Credit Cards and Top-Ups

By Nicho

Yes, the DeepSeek API takes PayPal. DeepSeek's help center lists PayPal, bank cards, Alipay and WeChat Pay for top-ups on platform.deepseek.com, and it refunds unused balance. For a developer outside China, DeepSeek is one of the easier Chinese model APIs to pay for.

This guide covers sign-up, top-ups, refunds and invoices as they stood on October 28, 2026, checked against DeepSeek's official docs, help center and terms. It also covers when calling DeepSeek-V4.1-Flash through cheapaimodel makes sense, and when it doesn't. For cost math with peak and off-peak rates, see DeepSeek API pricing; for error codes, see DeepSeek API not working?.

Quick answers

  • Payment methods: PayPal, bank card, Alipay and WeChat Pay. The help center doesn't name card networks.
  • Minimum top-up: not stated in DeepSeek's public docs.
  • Expiry: top-up balances don't expire. Promotional credit can, and it is spent first.
  • Refunds: unused balance can be refunded from Billing → Refunds, minus any handling fees.
  • Invoices: request them under Billing → Invoices.

Does the DeepSeek API accept PayPal or credit cards?

The help center says you can top up "via PayPal, bank card, Alipay, or WeChat Pay on the 「Top Up」 page." It says "bank card" and doesn't list card networks, so whether a particular credit card goes through isn't stated. PayPal is the method it names first.

The API is prepaid. The English docs list prices in US dollars and the Chinese docs in yuan; DeepSeek's balance API reports each account's currency as either CNY or USD, and the docs don't say how it is chosen. DeepSeek's terms also say it may ask you to prove where your top-up funds come from.

Does DeepSeek balance expire? Refunds and invoices

  • Expiry: "Top-up balances do not expire." Granted balance, such as promotional credit, can expire, and it is used before your topped-up balance.
  • Refunds: "Unused balances can be refunded." For online payments you start the refund yourself, under Billing → Refunds. The English terms say the refund comes after deducting necessary costs such as handling fees.
  • Invoices: you apply under Billing → Invoices, and you can add usage details in the Remarks field.

DeepSeek API sign-up outside China

DeepSeek runs one platform for everyone, with an OpenAI-format base URL https://api.deepseek.com and an Anthropic-format one at https://api.deepseek.com/anthropic. The operator named in its terms is Hangzhou DeepSeek Artificial Intelligence Co., Ltd.

Its help center covers sign-in by email, Google, phone with an SMS code, and WeChat. Things worth knowing:

  • Email: some email domains are refused. DeepSeek recommends a major provider such as Gmail, Outlook, Hotmail or Yahoo.
  • "Current device environment error": you see this when DeepSeek flags your IP address or phone number as high risk. It suggests trying a different environment.
  • Identity verification: the help center has personal and enterprise identity verification flows. The English terms don't say when an overseas account needs it.

DeepSeek API rate limits and the 402 error

DeepSeek sets a concurrency limit per account, shared by all your keys: 2,500 concurrent requests for deepseek-flash and 500 for deepseek-v4-pro. Going over it returns HTTP 429. DeepSeek doesn't sell tiered plans, but it raises limits on request at no extra cost. An empty balance returns 402 Insufficient Balance. DeepSeek API not working? walks through the other errors.

DeepSeek peak and off-peak prices vs cheapaimodel

The model name deepseek-flash currently serves DeepSeek-V4.1-Flash. DeepSeek charges full price in peak hours, 01:00–04:00 and 06:00–10:00 UTC on weekdays (excluding Chinese public holidays), about a fifth of the hours in a week, and half price the rest of the time.

Per 1M tokens DeepSeek off-peak DeepSeek peak cheapaimodel
Input, cache miss $0.15 $0.30 $0.209351
Output $0.60 $1.20 $0.837404

Our prices were taken from cheapaimodel's live model list on October 28, 2026, and are the same at every hour. They are 30% below DeepSeek's peak price, but above its off-peak price. If you only use DeepSeek and most of your traffic runs off-peak, buying from DeepSeek directly will usually cost less. DeepSeek API pricing works through example bills.

When cheapaimodel is worth it

  • Several model makers: you also call Kimi or GLM, and you want one account, one balance and one key for all of them.
  • Peak-hour traffic: most of your traffic falls in DeepSeek's peak window, and you want one predictable price.
  • Card-wallet payment: you prefer a credit or debit card (Visa, Mastercard, American Express, JCB), Apple Pay, Google Pay or Link through Stripe Checkout, in US dollars. The minimum top-up is $5 and there's no subscription.
  • Per-key controls: you want a spending limit, an expiration date, and model or IP allowlists on each key.

To weigh DeepSeek-V4.1-Flash against the other models on the same key, see DeepSeek-V4.1-Flash vs Kimi K3 and DeepSeek-V4.1-Flash vs GLM-5.3.

When DeepSeek direct is the better fit

  • Off-peak traffic: most of your traffic runs off-peak.
  • DeepSeek-V4-Pro or the Anthropic format: you need deepseek-v4-pro, or the Anthropic-format endpoint. We don't offer either.
  • Refunds: you want to be able to refund unused balance. cheapaimodel top-ups are non-refundable except as set out in our terms.

Call DeepSeek-V4.1-Flash through cheapaimodel

Use any OpenAI SDK. Set the base URL to https://api.cheapaimodel.ai/v1 and the model to deepseek-v4-1-flash:

from openai import OpenAI

client = OpenAI(
    base_url="https://api.cheapaimodel.ai/v1",
    api_key="YOUR_API_KEY",
)

reply = client.chat.completions.create(
    model="deepseek-v4-1-flash",
    messages=[{"role": "user", "content": "Explain the difference between a mutex and a semaphore."}],
)
print(reply.choices[0].message.content)

The quickstart walks through sign-in, top-up and your first key. Live prices for every model are on the pricing page.

Using Kimi or GLM too?

This is one of three guides for developers outside China. The other two cover the Kimi API and the GLM API on Z.ai. All of these models run on one cheapaimodel key; see all open-source models.

Sources: DeepSeek API docs, DeepSeek pricing, DeepSeek rate limit, DeepSeek error codes, DeepSeek help center: API, DeepSeek help center: login, DeepSeek Open Platform terms.

All posts