Claude Code is the only supported model here — and it’s a pre-release version labeled “GPT-5.x,” which doesn’t exist anywhere else. This is not a general-purpose relay. Hajimi (hajimi.ai) is a niche station for developers who want early access to Anthropic’s Claude Code tool, specifically the bleeding-edge “GPT-5.x” build. The station is new, with a 94% uptime and a safety rating of 3/5 — that’s mediocre, not dangerous, but expect occasional rough edges.
Pricing & Access
The station is currently free to try. No monthly subscription, no minimum recharge requirement. You can pay via Alipay or WeChat Pay if you decide to top up later, but the pricing model beyond the free trial is unspecified. No promo code is available.
| Plan | Price | Details |
|---|---|---|
| Free Trial | ¥0 | Full access to Claude Code (GPT-5.x) |
| Post-Trial | Not specified | No published recharge minimum |
| This makes it a zero-risk test for anyone curious about Claude Code. The catch: reliability is unproven. Only 56.4/100 composite score, and there’s no refund policy published. |
Model Support
- Claude Code (GPT-5.x) — This is the only model explicitly listed. Do not expect GPT-4, Claude 3.5, or any other model. The “GPT-5.x” label is unusual — it’s not a standard OpenAI or Anthropic model name. Likely a custom build or internal codename. Use with caution.
- No other models — If you need GPT-4, Claude 3.5 Sonnet, or Gemini, skip this station. This is a single-purpose relay.
China Access & API Compatibility
The station works without VPN — that’s the baseline for this review site. Payment via Alipay and WeChat Pay confirms domestic Chinese user focus. API compatibility is not documented, but since it’s a single-model relay, expect a simple API endpoint for Claude Code.
Uptime & Reliability
94% uptime is below industry standard. Most relay stations aim for 99%+. For a new station, this is acceptable but not great. Combined with the 3/5 safety rating, expect occasional downtime or degraded performance.
Pros & Cons
Pros
- Claude Code exclusive — unique access to a pre-release build
- Free to try with no upfront cost
- Accepts Alipay and WeChat Pay Cons
- Very limited public data and reviews — hard to verify claims
- Only one model supported — no fallback
- 94% uptime and 56.4 composite score indicate unproven reliability
- No refund policy or published pricing after trial
Verdict
Hajimi is for adventurous developers who want to kick the tires on Claude Code without paying upfront. It’s not a daily driver — the uptime and safety scores are too low for production use. If you’re building a side project or evaluating Claude Code for the first time, sign up for the free trial. If you need reliable API access to multiple models, look elsewhere. The “GPT-5.x” model name is suspicious. Without public documentation or community reviews, treat it as experimental. The station might pivot or shut down without notice.
FAQ
Q: Can I use GPT-4 or Claude 3.5 through Hajimi? A: No. The only supported model is Claude Code (GPT-5.x). There are no other models listed. Q: Is the free trial really free? A: Yes. No recharge or subscription is required to start. After the trial, pricing is not specified. Q: How reliable is the API? A: Uptime is 94%, and the composite score is 56.4/100. Expect occasional downtime. Not recommended for production workloads. Q: What payment methods are accepted? A: Alipay and WeChat Pay only. Q: Is there a refund policy? A: No refund policy is published. Consider this a risk if you recharge.
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.
