犀速AI
Measured metrics
- Overall score
- 8.0 / 10 ★★★★
- Stability band
- Not monitored
- Uptime
- Not monitored
- Response latency
- 4516 ms
- Model count
- 19
- Backend engine
- —
- Verification
- Pending review
- Last checked
- 2026-10-22
- Outbound clicks
- 13 clicks
Published rates & minimums
The figures below are published by the service itself on public channels. We only log them, and they are not our measurements — always confirm on the service's own billing page.
Site states: Claude at 0.3x, GPT / Gemini / Grok at 0.1x of official list price; ¥1 = 100,000 credits, never expiring
Billing model: Pay-as-you-go
A rate of 1.0 means official list price; 0.3 means you pay 30% of it. Not sure how the maths works out? Run the rate calculator →
About this service
Xisu AI exposes GPT, Claude, Gemini and Grok through both OpenAI-compatible and Anthropic-native protocols, covering 19 models. A single key and a single balance serve both: point base_url at the gateway for OpenAI SDKs and clients, or ANTHROPIC_BASE_URL for Claude Code, with no code changes. Its public rate card marks each model as a discount off official list price — Claude at 0.3x, GPT, Gemini and Grok at 0.1x. One yuan buys 100,000 credits which never expire, and there is no post-paid tier or credit line. Top-ups go through Alipay or WeChat from a ¥50 minimum. The console lets you issue several keys with individual budget caps, and every request log can be checked by request_id for model name, token count and cost. The site also publishes a self-test method for confirming that the model you called is the real one, backed by a refund promise if substitution is proven.
Strengths
- Dual-protocol support: one base_url change for OpenAI SDKs, one ANTHROPIC_BASE_URL change for Claude Code
- the rate card states the discount against official list price, so it can be compared directly
- credits never expire and there is no post-paid threshold
- multiple keys can each carry their own budget cap
- publishes a self-test method for model authenticity and promises a refund if substitution is verified
Watch out for
- No operating entity is disclosed, WHOIS is privacy-proxied and there is no ICP filing
- the homepage states a ¥50 minimum top-up while the top-up and refund guide suggests starting with a ¥10 package
- prices at 0.1-0.3x of official list cannot come from official API billing, so the supply is pooled-account based and an upstream ban can interrupt service
Our take: A good fit for anyone who needs both OpenAI and Anthropic protocols without maintaining two sets of keys; Claude Code setup is the least work in this category. At 0.1-0.3x of official list price, however, that pricing cannot come from official API billing — stability tracks the ban cycle of the upstream account pool. Do not route a critical production path through this channel alone; buy the smallest package, measure it on your own traffic, then scale.
Model families supported
Show the 19 models the endpoint probe detected (first 10)
claude-sonnet-5-5claude-opus-5-5claude-fable-5-1claude-haiku-5-5gpt-6.1-solgpt-6-astragpt-5.6-solgemini-4-argon-highgemini-3.8-flash-highgrok-4.7 Current promotions
No public sign-up bonus is listed; the vendor offers a free test key on request by email. Telegram support: @xisuai_support
Promotions are announced by the service itself and may change or end at any time — check its own page before relying on them.
Daily monitoring log
| Date | Uptime | Avg latency | Samples | Status |
|---|---|---|---|---|
| 2026-10-27 | 100.00% | 4516 ms | 1 | 200 |
| 2026-10-26 | 100.00% | 2164 ms | 2 | 200 |
| 2026-10-25 | 100.00% | 2002 ms | 2 | 200 |
| 2026-10-24 | 100.00% | 2269 ms | 2 | 200 |
| 2026-10-23 | 100.00% | 2041 ms | 2 | 200 |
Data sources
站点公开页面与运营方邮件提交（2026-10-22）+ 本站接口实测
We are not affiliated with 犀速AI and do not act as its agent, and we do not vouch for its service or the safety of your funds. Figures may lag or contain errors — if you spot one, please report it.
Similar services
Avian
Per the service's own site: Fast AI inference billed per token. DeepSeek V3.2, Kimi K2.5, GLM-5.…
apipro.maynor1024.live
An overseas AI API relay and aggregation service. Model coverage spans Claude, GPT, Gemini. It e…
api.cursorai.art
A globally hosted relay that fronts a range of AI model APIs. Claude, GPT and other families are…
LLM Gateway
The site itself puts it this way: Route, manage, and analyze LLM requests across OpenAI, Anthrop…
MegaLLM
Per the service's own site: The reliable AI gateway for developers. Access GPT-5, Claude Opus 4.…
Requesty
Its own site describes it as: Route every LLM call through one OpenAI-compatible API. Claude, GP…