Chart · Our World in Data
ChartOur World in Data
Training computation vs. dataset size in notable AI systems
Computation is measured in total petaFLOP, which is 10¹⁵ floating-point operations estimated from AI literature, albeit with some uncertainty. Training dataset size refers to the number of unique data points used to train the model.
Copy actions record publisher intent only. They do not confirm publication, citation or a backlink.
Data visualization loaded directly from Our World in Data.
Reuse conditions
Usage rights
- Source-hosted embed use
- Commercial use
- Modification
- Attribution
- Citation
- Raw-data redistribution
Chart and embed permission is separate from raw-data redistribution. The raw dataset is not included in this marketplace asset unless that right is explicitly marked allowed.
View rights evidencePublisher workflow
Exact citation
Epoch AI (2026)
Attribution: Epoch AI
Source-hosted only
Embed instructions
<iframe src="https://citesupply.com/e/training-computation-vs-dataset-size-in-notable-ai-systems" title="Training computation vs. dataset size in notable AI systems" loading="lazy" referrerpolicy="strict-origin-when-cross-origin" sandbox="allow-scripts" style="width:100%;height:720px;border:0;display:block"></iframe>
Embeds stay hosted by the source. Cite Supply records an aggregate iframe-load signal before redirecting to the reviewed source embed; it does not proxy or republish the chart.