2.1 seconds
average generation time per molecule
As published on nebius.com. Captured by usedby on Oct 7, 2026.
What happened
Quantori's cheminformatics team trains an equivariant diffusion model on a Nebius GPU cluster to generate molecules whose 3D shapes resemble a reference structure. A pre-trained Structure Seer model then predicts the bonds for the generated atom coordinates.
Summary written by usedby from the source page, in English. The figures are those of Nebius AI Cloud and Quantori, not ours.
- 1.6 million“the GPU capacity allowed training on a dataset containing 1.6 million compounds”
- 13.5 days“After 1,500 epochs (about 13.5 days) of training”
- 8 NVIDIA H200 GPUs“using a cluster on the Nebius cloud equipped with 8 NVIDIA H200 GPUs”
The quick provisioning and easy configuration of these resources enabled Quantori to set up training and evaluation with minimal effort, while the GPU capacity allowed training on a dataset containing 1.6 million compounds, ultimately leading to better model quality.
What the story claims, and what we checked
We compared the story with its live page on Oct 7, 2026.
- The figure: 2.1 secondsCheckedThe figure is on the page; its label is our wording.
- The passage quoted aboveCheckedCopied word for word from the page, near the name of Quantori.
- The numbers in our summaryCheckedEach one is printed on the page.
- Quantori uses Nebius AI CloudCheckedConfirmed line. Latest check across sources: Oct 7, 2026.
- The result itselfNot checkedWe quote it; we did not measure it.




