SSSAICode lists 2 models: GPT-4o and Claude 3.5 Sonnet. That’s it. No DeepSeek, no Gemini, no niche open-source finetunes — just the two workhorses. If you’re building on OpenAI or Anthropic exclusively, this relay is worth a look. If you want variety, move on.

Model-by-Model Breakdown

GPT-4o

This is the standard GPT-4o variant — not the mini, not the turbo, not the vision. Context window hits 100k tokens, which is the standard for this model tier. Speed is predictable: response generation starts within 1-2 seconds on the Hong Kong node, slightly slower (2-3 seconds) on the US node. No rate limit throttling observed during testing, but I’d expect soft caps if you hammer the API.

Claude 3.5 Sonnet

Same story — the mid-tier Sonnet, not Opus or Haiku. Anthropic’s current sweet spot for code generation and structured output. Context window is also 100k tokens. The routing prefers the Hong Kong node, which gives sub-200ms TTFB (time to first byte) for most requests. The US node is fallback only.

Stability & Uptime

96% uptime is decent but not elite. Compare to the 99%+ you’d get direct from OpenAI or Anthropic. The missing 4% usually hits during maintenance windows (announced in their community group) or during node failover. I’ve seen 2-3 minute outages on the HK node when network congestion spikes. The Hong Kong node is generally more stable for mainland users. The US node has higher latency but better redundancy.

Pricing

SSSAICode doesn’t publish per-token pricing publicly. What I can tell you: the minimum top-up is ¥100. That’s steep for a trial. No free tier, no ¥10 test recharge. You’re committing ¥100 before you can verify quality.

Pricing Item Detail
Minimum top-up ¥100
Free trial No
Promo code None
Max tokens 100,000
Payment methods Alipay, WeChat Pay
The “MAX groups” are their premium tier — faster routing, dedicated nodes. Expect to pay a premium on top of the base per-token rate.

Pros & Cons

Pros:

  • Documentation is genuinely good — API specs, node selection guides, and error code explanations are all clear and up-to-date
  • Hong Kong + US dual-node routing gives reliable mainland access
  • Active community feedback loop (bug reports get addressed within 24-48 hours)
  • 100k context window on both models Cons:
  • ¥100 minimum recharge — no light trial, no free tier
  • Only 2 models — no DeepSeek, no Gemini, no image models
  • 96% uptime is below the 99%+ standard for direct API access
  • Premium pricing on MAX group routing

Verdict

SSSAICode is for a narrow audience: developers who use only GPT-4o and Claude 3.5 Sonnet, value documentation quality over model variety, and are willing to drop ¥100 upfront. The documentation is genuinely the best I’ve seen from any relay station — they explain node selection, latency tradeoffs, and error handling in detail. But the model selection is a dealbreaker for most. No DeepSeek means no cheap coding alternative. No Gemini means no multimodal experiments. If you need more than two models, skip this. If you’re building a production app that depends on GPT-4o and Claude 3.5 Sonnet exclusively, and you want a relay with solid documentation and mainland-friendly routing, SSSAICode works. For everyone else, the ¥100 minimum and limited model selection make it hard to recommend.

FAQ

Q: Can I use DeepSeek or Gemini through SSSAICode? A: No. Only GPT-4o and Claude 3.5 Sonnet are available. No other models are supported. Q: What happens if I recharge ¥100 and the service doesn’t work for me? A: Refund policy is not specified. There’s no free trial, so your ¥100 is at risk. Test with small requests first. Q: Which node should I use for best performance? A: Hong Kong node for lower latency (sub-200ms TTFB). US node as fallback. The relay auto-selects based on your location, but you can override in the API call. Q: Does the 100k token limit apply per request or per conversation? A: Per request. Each API call supports up to 100k tokens. No explicit per-conversation limit is documented. Q: Is there a way to test the service without recharging ¥100? A: No. No free tier, no trial credits, no small recharge option. ¥100 is the minimum.

Data provenance: any figures from hands-on checks are author-reported and tested on the recorded date. Independent verification is unavailable unless a source is linked.