SUPER II: the model and the computing platform
The SUPER II platform is built by the Super Intellect team. A core on frontier-class models, surrounded by planning, result checks, tools and dedicated work environments for AI employees.
of video memory is needed for one working copy of a model of this class. That is more than 2 TB
is consumed by the accelerator cluster that runs such models
the cluster runs around the clock with active load and cooling management
Computing platform specs are given for the class of models SUPER II runs on.
Why SUPER II is paid
A model of this class does not fit into any single computer. It needs an accelerator cluster, electricity, cooling and people who look after it. These are real costs every hour.
The model does not fit into one graphics card
- Powerful gaming graphics cardup to 32 GB
- NVIDIA H200 class accelerator141 GB
- NVIDIA B300 class accelerator288 GB
- One working copy of the modelabout 2100 GB
That is about 15 H200 class accelerators just to hold one copy of the model in memory. Plus servers, network, power and cooling around them.
What the price of an AI employee is made of
- 01
Computing
Accelerators of the NVIDIA H200, B200 and B300 class. Cluster capacity is shared between client tasks.
- 02
Power and cooling
Tens of kilowatts around the clock, active cooling and load management.
- 03
Dedicated environment
Every paid employee has its own Windows machine with a browser and programs.
- 04
Maintenance
Updates, monitoring, security and support during business hours.
That is why the price of an AI employee is close to the cost of computing and maintenance. You pay not for chat access, but for working capacity that does your tasks 24/7.
A breakthrough in how business works with AI
AI should not just answer in a chat, it should work: take tasks, act in programs and report on the result. The Super Intellect team builds the SUPER II platform for this.
AI teams
Role-based employees who work together like a department.
Dedicated environments
Every paid employee has its own Windows environment: browser, programs, files.
Swarms
A big task is split between several employees and done in parallel.
Autonomous work
The employee plans its own steps and works 24/7, and agrees important actions with you.
The path of a task through SUPER II
The core thinks and makes decisions. Fast models do the small work, tools provide facts and files. Everything is checked before the answer.
- 1
Request
You describe the task in words, as you would to an employee.
- 2
SUPER II plans
Splits the task into steps and chooses who does each one.
- 3
Model cascade
Micro tasks go to fast compact models.
- 4
Tools
Facts from the web, the knowledge base and the work environment. Calculations go through a calculation tool.
- 5
Check and result
SUPER II checks the work and returns a finished result: a table, a file or a report.
An AI employee works like a person at a computer
Not only through APIs, but in a real Windows environment with a browser and programs. You see the screen and can stop the work at any moment.
- Every AI employee works in its own dedicated Windows environment
- Opens websites, accounts and programs the way a person does
- The environment screen is visible in your dashboard: screenshots every few seconds
- Sends emails, publishes and pays only after your confirmation
- Separate browser profiles for working with several accounts
- Files and access stay inside the environment and are not mixed with other clients
SUPER II for robotics
An AI core for robots and automated systems: warehouses, manufacturing, service robots, drones and smart buildings.
Learn moreAbout the model and the platform
Why is SUPER II paid?
Is it your own model?
Why not one big model for everything?
What does the free plan run on?
Is my data used for training?
Why does a tool do the math instead of the model?
Build your AI team
Free plan with no card: 1 employee in chat. Dedicated environment on paid plans.