Mars HK 是位于香港端口 31328 的 API 中转服务，通过 OpenAI 兼容接口提供 AI 语言模型访问，为亚太用户提供低延迟服务。
http://mars-hk.duckdns.orghttps://mars-hk.duckdns.org:31328完成站点认证，解锁以下权益：
https://mars-hk.duckdns.org:13111排名基于社区提交的测试数据与定期健康探测，仅供参考，非官方数据。
Mars HK appears to expose OpenAI-compatible API endpoints from a Hong Kong-hosted relay service. The public website currently returns a 404 page, so this entry is documented conservatively from the reachable domain, existing sibling metadata, and observed endpoint behavior.
按 13 个模型行对比最新 audit、最新测速、吞吐、延迟与按 token 计费价格。
| 模型 | 输入 ($/M) | 输出 ($/M) | 检测 | 速度 | 延迟 |
|---|---|---|---|---|---|
| - | - | — | 54.3 t/s | 6.62 s | |
codex-auto-review | - | - | — | 49.7 t/s | 2.15 s |
gpt-5.5 | - | - | — | 30.1 t/s | 6.53 s |
deepseek-v4-prodefault | $75.00/M | $75.00/M | — | — | — |
deepseek-v4-flashdefault | $75.00/M | $75.00/M | — | — | — |
glm-5.1default | $75.00/M | $75.00/M | — | — | — |
MiniMax-M2.7default | $75.00/M | $75.00/M | — | — | — |
minimaxai/minimax-m2.7default | $75.00/M | $75.00/M | — | — | — |
gpt-5.4-minidefault | $75.00/M | $450.00/M | — | 128.3 t/s | 4.45 s |
glm-5-turbodefault | $75.00/M | $75.00/M | — | — | — |
gpt-5.4default | $75.00/M | $450.00/M | — | 48.5 t/s | 6.08 s |
gpt-5.4(high)default | $75.00/M | $450.00/M | — | — | — |
gpt-5.4(xhigh)default | $75.00/M | $450.00/M | — | — | — |
当前显示 13 / 13 个模型行
| 时间 | 模型 | 速度 | 延迟 |
|---|---|---|---|
| May 9, 01:30 AM | gpt-5.4-mini | 148.52 tok/s | 3.07s |
| May 9, 01:29 AM | gpt-5.4-mini | 135.95 tok/s | 6.05s |
| May 6, 04:34 AM | gpt-5.5 | 59.52 tok/s | 3.94s |
| Apr 28, 01:30 PM | MiniMax-M2.7-highspeed | 54.34 tok/s | 6.62s |
| Apr 26, 06:13 AM | gpt-5.5 | 29.21 tok/s | 16.15s |
| Apr 26, 06:03 AM | codex-auto-review | 49.68 tok/s | 2.15s |
| Apr 24, 02:33 PM | gpt-5.5 | 31.97 tok/s | 6.97s |
| Apr 24, 01:59 PM | gpt-5.5 | 17.28 tok/s | 6.21s |
| Apr 24, 01:59 PM | gpt-5.4-mini | 100.55 tok/s | 4.24s |
| Apr 24, 01:57 PM | gpt-5.4 | 53.62 tok/s | 8.01s |
用 179 个 LMSpeed 信号，把 Mars HK 的 6 个相近 API 替代服务商放在一起比较：共享模型覆盖、价格、实测速度、可用性和免费模型。
| 服务商 | 对比理由 | 模型数 | 免费项 | 均价 | 速度 | 30 天可用性 |
|---|---|---|---|---|---|---|
| Mars HK mars-hk-duckdns-org-31328 Mars HK is an API relay service hosted in Hong Kong on port 31328, providing access to AI language models through OpenAI-compatible endpoints with low latency for Asia-Pacific users. | 当前服务商基线 | 22 | 42 | N/A | 52 tok/s | 0.7% |
| CaMeL AI api-kr777-top CaMeL AI provides an OpenAI-compatible API gateway with extensive model coverage and pricing options. |
| 194 | 4 | $88.30/M | 93 tok/s | 99.1% |
api-vectorengine-ai Vector Engine provides an API platform aggregating access to over 500 AI large models with OpenAI API compatibility and global deployment. |
| 182 | 6 | $5.75/M | 89 tok/s | 99.6% |
api-n1n-ai N1N provides API access to a wide range of AI models including GPT-4, Claude 3, Gemini, and others for text, image, and video generation. |
| 177 | 6 | $8.65/M | 90 tok/s | 99.7% |
api-futureppo-top Future Hub is a large-scale OpenAI-compatible API gateway at api.futureppo.top with extensive model coverage and pricing data. |
| 132 | 223 | N/A | 263 tok/s | 0.6% |
180txt-cn 180txt API provides an OpenAI-compatible API relay for multiple AI models. |
| 37 | 77 | $18.03/M | 42 tok/s | 0.7% |
x666-me X666 API is an OpenAI-compatible API relay offering access to multiple AI models with competitive pricing and high availability. |
| 25 | 0 | $0.0014/M | 77 tok/s | 99.3% |
