Public Geekbench 7 results attributed to OpenAI’s dots suggest a virtual machine with nine AMD EPYC 9V74 cores and 9.73 GB of memory. In Tom’s Hardware’s reported sample, six post-launch dots runs had a multi-core median of about 8,550—roughly six times Meta Muse’s median of about 1,394. That gap chiefly compares differently provisioned virtual machines; it is not evidence that dots is six times better or faster at agent tasks.
What the Geekbench results show
Tom’s Hardware reporter Shane Downing said on September 30, 2026, that a check of Geekbench 7’s public database found six post-launch runs associated with dots on the reported nine-core configuration. The results ranged from 1,512 to 1,614 in single-core testing and from 8,135 to 8,991 in multi-core testing. The reported medians were 1,570 single-core and about 8,550 multi-core.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
AMD Epyc 9554 Processor 3.1 Ghz 256 Mb L3, W128281619 (256 Mb L3) | $3,734.45 | Buy on Amazon |
| 2 |
|
AMD Epyc 9354 Processor 3.25 Ghz 256 Mb L3, W128281623 (256 Mb L3) | $2,819.95 | Buy on Amazon |
| 3 |
|
AMD EPYC 9004 [4th Gen] 9124 Hexadeca-core [16 Core] 3 GHz Processor | $977.48 | Buy on Amazon |
As an Amazon Associate I earn from qualifying purchases.
There was also a higher pre-launch result: 9,435 multi-core points. According to Tom’s Hardware, the screenshot was uploaded on September 25, four days before dots launched, and the person who ran it is unknown. That score is nearly 5% above the best of the six post-launch runs and about 10% above their median. It may be an outlier, though it falls in the same broad performance range.
Recommended Free Tools
These figures are reported from public benchmark uploads, not an official OpenAI hardware disclosure or a controlled comparison of agent performance. Tom’s Hardware’s report describes the database check and its results.
#1 Best Overall
- Sockel SP5, 64 x 3.1 GHz (Boost 3.75) GHz
- 384 MB L3 Cache, 64 cores/ 128 threats
- 12-channel memory support up to DDR5-4800 MHz
- Max. Performance consumption 360 watts (structural width 5 Nm)
- Tray (without cooler)
How dots compares with Meta Muse
In the same report, ten Geekbench 7 runs attributed to Meta Muse had medians of about 1,041 single-core and 1,394 multi-core. An independent Muse run cited by the article scored 1,000 single-core and 1,433 multi-core.
| Reported Geekbench 7 results | Dots | Meta Muse |
|---|---|---|
| Runs in the reported sample | 6 post-launch runs | 10 runs |
| Single-core median | 1,570 | About 1,041 |
| Multi-core median | About 8,550 | About 1,394 |
| Reported virtual-machine CPU | AMD EPYC 9V74 | AMD EPYC 9D25 |
| Reported cores and base clock | 9 cores; 2.6 GHz base clock | 2 cores; 1.5 GHz base clock |
| Reported memory | 9.73 GB | 7.75 GB |
In this sample, dots’ single-core median was about 1.5 times Muse’s, while its multi-core median was about six times as high. The nine-versus-two core allocation and different reported base clocks matter: multi-core scores reflect the amount of CPU capacity as well as processor characteristics. The report suggests clock speed and architecture may account for some of the remaining difference, but the figures do not isolate those effects.
Rank #2
What CPU and memory does dots appear to use?
The benchmark-system description interpreted by Tom’s Hardware points to a nine-core slice of an AMD EPYC 9V74 virtual machine with 9.73 GB of memory. Those are details inferred from public benchmark records and secondary reporting, not specifications formally confirmed by OpenAI. The report does not establish whether the assigned cores remain reserved while a dot is idle.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesOpenAI described a dot as having “their own cloud computer, their own browser, and the apps you’ve connected.” Dots launched at DevDay for Pro and Business Premium users, according to the September 30 report; availability can change, so check OpenAI’s current product information for present eligibility. The report also relays one user’s account that a launch-day dot identified Debian Linux and listed apps including Chromium, Blender, GIMP, Inkscape, Kdenlive, Godot, FreeCAD, OpenSCAD, KiCad, QGIS, ParaView and 3D Slicer, plus Python, Node.js and Git. That account is not an official inventory of every dot’s software.
Rank #3
What Geekbench does—and does not—tell you
Geekbench CPU scores offer a snapshot of processor performance under the benchmark’s workloads. They do not directly measure how well an AI agent completes a user’s task, how quickly it responds end to end, how reliably it works, or what it costs to operate. The reported sixfold multi-core gap therefore describes these VM benchmark samples, not a sixfold difference in overall agent capability.
The result also cannot establish that every dots session receives the same machine allocation. The report discusses a possible future ability to add dots and increase their speed, but does not establish that scaling controls are already available.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
