In 2024 the biggest training run would still be about 5×10^25 FLOP
“We don’t see anything substantially bigger. Corps spend their money fine-tuning and distilling and playing around with their models, rather than training new or bigger ones. (So, the most compute spent on a single training run is something like 5×10^25 FLOPs.)”
Context: "What 2026 looks like" (LessWrong, Aug 2021), 2024 section.
How we judge it
Our interpretation: correct if the largest publicly known training run of models released in 2024 is within a factor of 2 of 5e25 FLOP.
Resolution criteria are our interpretation of the claim, written before grading. Methodology
- Said
- Resolve by
- Verdict
- Partly correct ()
- Stated probability
- None stated
- Specificity
- 3 of 3 (precise and measurable)
- Score
- 0.5 points × weight 3
Evidence & grader’s note
Partly: remarkably close for released models, but the claim that nothing bigger was being trained did not hold.
Evidence timeline
Dated entries from the claim’s own record (statement, verdict, later developments) and sourced entries added by editors. Each links to its source; none changes the verdict.
Share this scorecard
Post on X LinkedIn Share image (PNG)
Embed / Cite
The embedded card is live: it shows the current verdict, with the date the data was read. If a verdict is corrected later, the card changes too, so cite the page and its date for a fixed record.