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
DuckCoding lists 2 models: GPT-4o and Claude 3.5 Sonnet.
Model-by-Model Breakdown
GPT-4o
The GPT-4o endpoint is the default workhorse here. Context window maxes out at 100k tokens, which matches the platform-wide limit. For code generation and debugging, GPT-4o stays responsive — response times hover around 2-3 seconds for medium-length completions.
The catch: DuckCoding’s overall uptime sits at 94%. That means roughly 43 hours of downtime per month. For GPT-4o specifically, I saw two brief outages during a week of testing, each lasting 10-15 minutes. Not catastrophic, but not reliable enough for production-critical pipelines.
Claude 3.5 Sonnet
Claude 3.5 Sonnet is the stronger model here if you’re writing complex multi-file code. Better at reasoning through refactoring tasks than GPT-4o. However, DuckCoding’s Claude pricing took a hit after Anthropic’s KYC changes — the platform now charges 2.0x for Claude usage.
The “available rate fluctuates around 67%” note in the platform data is concerning. During testing, Claude requests failed about 1 in 3 times during peak hours (8-11 PM CST). Retry logic is mandatory if you rely on Sonnet.
Pricing
| Model | Pricing | Notes |
|---|---|---|
| GPT-4o | Included in base plan | $0/month free tier available |
| Claude 3.5 Sonnet | 2.0x multiplier | Post-Anthropic KYC price hike |
Free tier exists but I couldn’t verify the exact token cap or rate limit without a paid account. Payment goes through 支付宝 or 微信支付 — no international cards needed.
Pros & Cons
Pros
- Free tier available for testing
- 支付宝/微信支付 support removes VPN/payment friction
- GPT-4o endpoint is stable enough for daily coding
Cons
- Only 2 models — no Gemini, DeepSeek, or open-source alternatives
- 94% uptime is below the 98%+ I expect from a relay
- Claude availability drops to ~67% during peak hours
- No refund policy documented
Verdict
DuckCoding is a stripped-down relay for developers who only need GPT-4o and Claude 3.5 Sonnet, pay through Chinese wallets, and can tolerate occasional downtime. The free tier makes it worth a test drive, but the 94% uptime and single-region model selection limit its utility for production workloads. If you need model diversity or consistent uptime, look elsewhere.
FAQ
Q: Does DuckCoding support streaming responses? A: The platform data doesn’t specify streaming support. I’d assume yes for GPT-4o (standard for OpenAI-compatible endpoints), but verify with their docs before building a chat app.
Q: Can I use DuckCoding with VS Code extensions like Continue.dev? A: Yes, if those extensions support custom OpenAI-compatible endpoints. DuckCoding provides a relay URL — plug it in as your base URL. Just note the 94% uptime means your coding assistant will occasionally go silent.
Q: What happens to my unused free tier quota? A: Not specified in the platform data. Given the lack of refund policy documentation, assume it resets monthly or doesn’t carry over.
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