Dots run on 9-core AMD EPYC VMs: Geekbench data shows ~6x multi-core lead over Meta's Muse

#What happened

Geekbench 7 results uploaded before and after the DevDay launch show OpenAI's Dots agents operate on nine-core slices of AMD EPYC 9V74 processors with 9.73GB of memory. Six verified post-launch submissions score 8,135–8,991 multi-core and 1,512–1,614 single-core. A pre-launch benchmark shared on X by user INIYSA on Sept. 25 (four days before launch) hit 1,667 single-core / 9,435 multi-core — about 5% above the best post-launch run, an apparent outlier.

Meta's Muse, by comparison, runs on two-core EPYC 9D25 instances with 7.75GB of memory, posting median scores of 1,041 single-core and 1,394 multi-core across 10 database entries. That leaves Dots with a ~1.5x single-core and ~6x multi-core lead; the nine-vs-two core-count gap accounts for most of it, with 2.6GHz vs 1.5GHz base clocks playing a role.

On launch day, a user reported her dot's preloaded environment: Chromium, Blender, GIMP, Inkscape, Kdenlive, Godot, FreeCAD, OpenSCAD, KiCad, QGIS, ParaView, 3D Slicer, plus Python, Node.js and Git — running Debian Linux (the pre-launch benchmark ran on Ubuntu). OpenAI says users will eventually be able to add more dots and "scale the output of each dot by ... increasing its speed."

#Why it matters

The agent wars are turning into an infrastructure arms race: always-on agents need dedicated compute. Muse has reportedly passed 500,000 daily active users, and SpaceXAI plans standalone Nvidia Vera CPUs for Grok's agentic workloads. OpenAI bills dots as having "their own cloud computer, their own browser, and the apps you've connected." Whether dots hold their allocated cores while idle is still unknown — so the true cost of always-on agents remains unpriced.

Verified Oct 1, 2026.

SOURCES: https://news.lavx.hu/article/openai-s-dots-agents-run-on-nine-core-amd-epyc-vms-geekbench-data-shows-sixfold-multi-core-lead-over-meta-muse (LavX News)