I couldn't pay for DeepSeek V4 with my card. So I built a proxy

zftnb6661 pts0 comments

I built a DeepSeek API proxy because I couldn't pay with my credit card – api-hub.cc<br>A">

📅 July 12, 2026<br>📖 5 min read

I built a DeepSeek API proxy because I couldn't pay with my credit card

DeepSeek V4 is incredible value. But the official platform won't take my credit card. Here's how I fixed it.

The problem

DeepSeek V4-Flash is roughly $0.14 per million input tokens from the official API — a fraction of what OpenAI or Anthropic charge. It supports an OpenAI-compatible endpoint, so in theory you can drop it into any existing project.

There's just one catch: the official DeepSeek platform only accepts Alipay and WeChat Pay . If you're an international developer with a Visa or Mastercard, you literally cannot pay them.

I wanted to use DeepSeek for a project. I couldn't. So I built a proxy that accepts credit cards.

How it works

The stack is straightforward:

LiteLLM — routes requests to DeepSeek, manages API keys and usage tracking

Creem.io — payment processing (Visa, Mastercard, Apple Pay, Google Pay)

PostgreSQL — stores key metadata and usage data

Node.js/Express — auth, dashboard, key management

When you buy credits, the server calls LiteLLM's /key/generate to create an API key with a spending limit matching your purchase. The key works with any OpenAI-compatible SDK — just change base_url to https://api-hub.cc/v1 and you're set.

What DeepSeek V4-Flash offers

Price. Flash input is ~$0.14/M tokens officially. Through api-hub.cc it's $0.20/M — slightly above raw cost, but you get instant provisioning, a dashboard, and credit card payment.

Speed. Flash generates at 100+ tokens per second. In my experience it's faster than any Anthropic or OpenAI model I've used via API.

OpenAI compatibility. Same endpoint format, same SDK. Zero code changes needed.

Prompt caching. Cache hit pricing drops to $0.02/M for Flash — 10x cheaper than uncached input. Just keep your system prompt consistent.

An interesting development: peak pricing

In late June 2026, DeepSeek announced a peak/off-peak pricing model for V4 Flash. During Beijing peak hours (9:00-12:00 and 14:00-18:00), official prices double to ~$0.28/M input.

api-hub.cc charges a flat $0.20/M for Flash input at all hours — which means during peak times, the proxy is actually cheaper than going direct.

Current pricing

Flash: $0.20/M input, $0.40/M output (cache hit: $0.02/M)

Pro: $0.60/M input, $1.20/M output (cache hit: $0.15/M)

Three tiers: $5 (casual use), $20 (daily driver), $100 (team/power user).

Final thought

The only thing standing between most international developers and DeepSeek's pricing is a payment method. api-hub.cc solves that one problem — everything else (the speed, the quality, the OpenAI compatibility) is already there.

Try it: api-hub.cc — create an account, pay with your card, get an API key in seconds. Feedback welcome at [email protected].

deepseek flash input card openai proxy

Related Articles