The 30-second summary
+ What we liked
- Codex subscription \$100/month (=¥700) with \$60/day quota
- Competitive Codex pricing
- Team+ and Plus account pools available
− What we didn't
- Claude Max pricing rose to 2.0x after Anthropic KYC changes
- Codex packages suspended temporarily then re-listed at higher price
- Available rate fluctuates around 67%
In-depth review
GPT-4o input costs $2.50 per million tokens on DuckCoding, and Claude 3.5 Sonnet runs at $3.00 per million input tokens — both roughly 1.3x the direct API rate from OpenAI and Anthropic. That markup is the price of avoiding VPN headaches in China, but it’s not the whole story.
Pricing Breakdown
DuckCoding doesn’t publish a per-token rate card on their site — you discover the actual costs after depositing. Based on community reports and the platform’s own data, the effective rates land at approximately:
| Model | Input (per 1M tokens) | Output (per 1M tokens) | Direct API Comparison |
|---|---|---|---|
| GPT-4o | $2.50 | $10.00 | ~1.3x OpenAI direct |
| Claude 3.5 Sonnet | $3.00 | $15.00 | ~1.3x Anthropic direct |
The free trial gives you limited tokens to test connectivity. No batch discounts exist — each request is billed individually at those per-token rates.
The Codex Angle
This platform specifically targets Codex users. A Codex subscription costs $100/month (≈¥700) with a $60/day quota. DuckCoding offers Team+ and Plus account pools that let multiple developers share that quota, which is the main appeal. If you’re solo and burning through $60/day on GPT-4o coding tasks, the math works: $100/month for $1,800/month worth of token usage.
But here’s the catch: Codex packages were temporarily suspended, then re-listed at higher prices. The available rate — meaning the percentage of time you can actually draw from the pool — fluctuates around 67%. That means roughly one out of three requests might fail or time out during peak hours.
Claude Pricing Volatility
After Anthropic’s KYC changes in late 2025, Claude Max pricing on DuckCoding rose to 2.0x the direct rate. If you were relying on Claude 3.5 Sonnet for coding, you’re now paying double what Anthropic charges directly — and you still can’t access Anthropic’s API from China without a VPN. The value proposition weakens significantly here.
Hidden Costs
Two things to watch:
- Minimum recharge isn’t specified. You might deposit ¥50 and discover the effective per-token rate is higher than expected because of rounding or minimum spend rules.
- No refund policy. If the service goes down (current uptime: 94%) or you decide to switch, your balance is stuck.
Pros & Cons
Pros
- Codex subscription at $100/month with $60/day quota is competitive for heavy GPT-4o users
- Team+ and Plus account pools reduce per-developer cost for teams
- 支付宝/微信支付 accepted — no foreign credit card needed
Cons
- Claude Max pricing jumped to 2.0x after Anthropic KYC changes
- Codex packages suspended and re-listed higher — trust is damaged
- Available rate fluctuates around 67% — expect one-third of requests to fail
- Uptime at 94% means ~22 hours of downtime per month
- No refund policy if you leave
Verdict
DuckCoding is a niche tool for one specific use case: Chinese developers who rely heavily on GPT-4o for Codex-style coding and need a team pool to share quota. The $100/month subscription with $60/day quota is genuinely cheaper than paying per-token at 1.3x markup for heavy users.
For everyone else — especially Claude users — the 2.0x markup and 67% availability rate make this hard to recommend. The 94% uptime and suspended-then-relisted packages signal operational instability. If you’re a solo developer using GPT-4o casually, you’re better off with a relay station that has 99%+ uptime and transparent per-token pricing.
FAQ
Does DuckCoding support batch API requests or rate limit pooling?
No batch discounts are offered. Each request is billed individually at the per-token rates. The Team+ and Plus account pools share a daily quota ($60/day for Codex), but individual rate limits aren’t documented.
What happens if my balance runs out mid-request?
The platform cuts off the request immediately. There’s no overdraft or grace period. Since minimum recharge isn’t specified, you’ll need to deposit more before continuing.
Can I use DuckCoding for production workloads with the 67% available rate?
Not reliably. A 67% available rate means roughly one-third of requests may fail during peak hours. Combined with 94% uptime, this isn’t production-grade. Use it for personal coding experiments or non-critical tasks.
Pricing breakdown
DuckCoding offers competitive pricing for developers. Here's the breakdown:
| Plan | Price | Quota | Best for |
|---|---|---|---|
| Free | $0/mo | Free trial | Kicking the tires |
| Standard RECOMMENDED | Pay-as-you-go/mo | Unlimited usage | Solo devs · small teams |
| Enterprise | Custom | SLA · dedicated support | Teams & agencies |
Supported models
2 models across major vendors.
Frequently asked questions
Can I access this platform from China without a VPN?
Most relay stations are accessible from Chinese ISPs. Check our review for specific routing details.
What payment methods are accepted?
Payment options vary by platform. Some accept Alipay/WeChat Pay, others are USD/crypto only.
How does this compare to using OpenAI directly?
Relay stations add routing latency but provide access from restricted regions, unified billing, and multi-model fallback.
Is my API key safe?
Keys are encrypted at rest. Most platforms support per-project scoping and IP allow-lists.
Should you use DuckCoding?
Codex users willing to pay for reliability