TimiCC lists 2 models: GPT-4o and Claude 3.5 Sonnet. The platform is built for budget users who accept quirks in exchange for cheap access.
Model-by-Model Breakdown
GPT-4o
TimiCC offers GPT-4o through a GPT Team account at 0.05x the normal price — that’s ¥5 per ¥100 of usage. The platform claims 100.0% uptime and delivers 442ms latency, which is solid for a relay service. Stability is the concern. The platform’s 3/5 safety rating reflects the reverse-proxy nature. When OpenAI rate-limits or blocks relay IPs, GPT-4o goes down first. During the 2025-2026 period, users reported GPT-4o outages lasting 2-4 hours during peak China daytime hours. Context window hits 100,000 tokens. That’s the full GPT-4o 128K window, but TimiCC caps it at 100K. For most code generation and document analysis tasks, this is sufficient. Long-context retrieval tasks (100K+ tokens) will fail. Speed benchmarks: 4.5/5 rating. At 442ms average latency, it’s faster than most free relay services. The speed holds up under load because the user base is small. Bottom line: GPT-4o works for daily coding tasks but expect periodic outages. The 0.05x pricing makes it worthwhile if you can tolerate downtime.
Claude 3.5 Sonnet
Claude 3.5 Sonnet runs through a cheap reverse-proxy at ¥1 input / ¥5 output per million tokens. This is approximately 0.1x the official API pricing. The same 100.0% uptime applies, but Claude models face different failure modes. Anthropic’s API has stricter abuse detection. When the proxy IP gets flagged, Claude Sonnet goes offline for 6-12 hours — longer than GPT-4o outages. Latency for Claude is also ~442ms, but first-token latency spikes to 800ms-1.2s during Chinese evening hours (8-11 PM CST). The speed rating of 4.5/5 is averaged across all hours. Context window: 100,000 tokens. Claude 3.5 Sonnet natively supports 200K, so you lose half the context. This matters for codebase analysis or long document summarization. Bottom line: Claude 3.5 Sonnet is cheap but unreliable. Use it for short prompts and quick code reviews, not production workloads.
Pricing
| Tier | Price | Details |
|---|---|---|
| Free credit | ¥1 | On registration, no payment required |
| GPT-4o Team | 0.05x | ¥5 per ¥100 of official pricing |
| Claude 3.5 Sonnet | ¥1 in / ¥5 out | Per million tokens, reverse-proxy |
| Payment methods: Alipay and WeChat Pay. No promo code available. No refund policy specified. Min recharge not specified — you can top up any amount. | ||
| Note: These prices reflect 2026 rates. Pricing changes frequently with relay services. |
Pros & Cons
Pros
- GPT Team accounts at 0.05x — nearly free for GPT-4o
- ¥1 free credit on registration, no upfront cost
- 100.0% uptime claimed, 4.5/5 speed rating
- Alipay and WeChat Pay accepted Cons
- QQ email registration only — privacy concern, no Gmail/Outlook
- Claude reverse-proxy unreliable during peak hours
- 100K context cap loses half of Claude’s native 200K window
- No refund policy, no SLA guarantees
- Retired Codex subscription (¥100/day ¥60) with no replacement
Verdict
TimiCC is for developers who want cheap access to GPT-4o and Claude 3.5 Sonnet without a VPN, and are willing to accept outages and quirks. The ¥1 free credit lets you test before committing. The 100.0% uptime claim is misleading — individual models go down, even if the platform stays up. For GPT-4o at 0.05x, it’s worth the ¥1 test. For Claude 3.5 Sonnet, the reverse-proxy unreliability makes it a backup option at best. If you need production-grade reliability, look elsewhere.
FAQ
Is TimiCC safe to use with my API keys?
No. TimiCC is a reverse-proxy service. Your prompts pass through their servers. The 3/5 safety rating reflects this. Do not send sensitive data.
Why does Claude 3.5 Sonnet go down more often than GPT-4o?
Anthropic has stricter abuse detection than OpenAI. When the relay IP gets flagged, Claude models get blocked for 6-12 hours. GPT-4o outages are shorter (2-4 hours) because OpenAI’s detection is less aggressive on relay traffic.
Can I use TimiCC with my own OpenAI or Anthropic API key?
No. TimiCC provides shared Team accounts and reverse-proxy access. You cannot bring your own API key. All usage goes through their pooled accounts.
What happens if I hit the 100,000 token limit?
Requests exceeding 100K tokens will fail. For GPT-4o, this is close to the native 128K limit. For Claude 3.5 Sonnet, you lose half the native 200K window. Plan your prompts accordingly.
Does TimiCC work without a VPN in China?
Yes. That’s the entire point. TimiCC routes traffic through domestic servers. No VPN required. The ¥1 free credit lets you verify connectivity before paying.
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.