OpenAI’s dots agents may be software, but each runs on a virtual computer with a defined share of CPU, memory and storage. The reported setup offers a useful glimpse of the resources behind each agent, though it does not tell us how well dots handles any particular task.
The machine behind dots
Each virtual machine runs on an AMD EPYC 9V74 server processor, an 80-core chip, and is assigned nine cores. The machine also has about 10 GB of RAM and 32 GB of storage.
Those resources are expected to cover many tasks for this kind of agent, though the available details leave open where dots might hit its limits. A virtual machine is still a computer with a finite allocation, even when someone else owns the server.
What the benchmark shows
Reported Geekbench 7 results for the virtual machines are 1,667 points in single-core testing and 9,435 in multi-core testing. These scores help describe the hardware setup, but they are not a direct measure of an agent’s ability to complete a particular job.
The figures sketch the boundaries of dots’ computing environment. They do not, on their own, explain how OpenAI divides work among agents or what happens when a task exceeds the resources available to one machine.