- Third-Party Servers第三方服务器
- Per-Token Bill按量计费
- Leak Risk泄露风险
- Your Own Sign-In沿用您的登录系统Microsoft Entra, Google Workspace or LDAPMicrosoft Entra、Google Workspace 或 LDAP
- Roles, Not People按角色授权Permissions attach to roles权限绑定角色,而非个人
- Department Scope部门范围A query searches only its own collection查询只检索本部门资料
- Every Action Logged全程记录Who, when, and what was retrieved谁、何时、检索了什么
- No Outbound Route无外连通道No path to the internet by default默认不连接互联网
- Encrypted全程加密TLS on the network, full-disk at rest传输 TLS 加密,存储全盘加密
8monthsto pay back即可回本
RM 58,450saved by month 24第 24 个月累计节省
20 staff on AI seats at RM 180 a month each, against a RM 27,000 build plus electricity. Worked example, not a quote.20 名员工每人每月 RM 180 的 AI 订阅,对比 RM 27,000 的服务器加电费。仅为示例,并非报价。
-
Ask提问
Staff use chat, files and apps from any office device员工在任何办公设备上使用对话、文件与应用
- Laptop笔记本电脑
- Phone手机
- Tablet平板
-
Think运算
The model runs on your server, inside your network模型运行在您的服务器上,位于内部网络
- Local Models本地模型
- Private Network私有网络
-
Answer回答
The reply returns to the same screen, nothing goes out结果回到同一屏幕,没有任何数据外传
- No Cloud无云端
- No Third Party无第三方
Model Size模型规模GPU Memory Needed所需显存Runs On运行设备
The 48 GB Tower48 GB 塔式服务器
7-8B
5-6 GB
An 8 GB laptop GPU8 GB 笔记本显卡
24-32B
16-20 GB
A 24-32 GB desktop GPU24-32 GB 台式机显卡
70B
40-43 GB
A 48 GB workstation GPU48 GB 工作站显卡
120B MoE
60-65 GB
A 96 GB workstation GPU96 GB 工作站显卡
671B
380-400 GB
Eight data-centre GPUs八张数据中心显卡
024487296 GB
4-bit models. Rule of thumb: GB = billions of parameters x 0.64 位量化模型。经验法则:显存 GB = 参数量(十亿)x 0.6
-
Desk AI · SoloDesk AI · 个人 Solo Professional个人专业用户- Installed
- Laptop, RTX 5050 8 GB笔记本电脑,RTX 5050 8 GB
- Used For
- Drafting, document Q&A, coding写作、文档问答、编程
- Fits
- One person一位使用者
-
Studio AI · TeamStudio AI · 团队 Multimedia University, Cyberjaya多媒体大学,赛城- Installed
- Tower, RTX PRO 5000 48 GB塔式服务器,RTX PRO 5000 48 GB
- Used For
- Team chat, document Q&A团队对话、文档问答
- Fits
- 20 to 50 people20 至 50 人
-
Engine AI · HeavyEngine AI · 重载 Hong Kong University香港大学- Installed
- 48U rack, HPC servers48U 机柜,HPC 高性能计算服务器
- Used For
- Research and AI training科研与 AI 训练
- Fits
- 60 to 150 people60 至 150 人
-
Rack AI · OrganisationRack AI · 机构 Multiple Racks多机柜- Installed
- Multi-rack cluster多机柜集群
- Used For
- Organisation-wide AI, agents全机构 AI、智能体
- Fits
- Whole organisation整个机构
Indicative prices, confirmed at quote. The rack is a VYROX photograph, the other images are illustrative renders价格仅供参考,以正式报价为准。机柜为 VYROX 实拍照片,其余图片为示意渲染图
12 m²fits one to four racks可容纳一至四个机柜
- 1 Cooling散热Precision aircon, N+1精密空调,N+1 冗余
- 2 Humidity湿度40 to 60% RH相对湿度 40 至 60%
- 3 Temperature温度18 to 27 °C18 至 27 °C
- 4 Access门禁Card or face, logged刷卡或人脸,全程记录
- 5 Power电力Three-phase and UPS三相电与 UPS
- 6 Fire Safety消防Clean-agent, no water洁净气体灭火,不用水
Desk and Studio AI need no room. Engine AI needs a closet. Rack AI needs this room.Desk 与 Studio AI 无需机房,Engine AI 一个机柜间即可,Rack AI 需要这样的机房。
1 / 12



