Skip to content
Longterm Wiki
Back

Epoch AI, "Frontier LLM training runs can't get much longer" (https://epoch.ai/data-insights/longest-training-run)

web

Credibility Rating

4/5
High(4)

High quality. Established institution or organization with editorial oversight and accountability.

Rating inherited from publication venue: Epoch AI

Relevant to forecasting AI progress and understanding scaling limits; informs debates about whether frontier capabilities can continue growing via training compute scaling alone.

Metadata

Importance: 55/100organizational reportanalysis

Summary

Epoch AI analyzes the physical and practical limits on how long frontier AI training runs can be extended, finding that training duration is approaching natural ceilings due to hardware reliability, data constraints, and optimization dynamics. The analysis suggests that simply scaling training time is not a viable path for continued capability gains at the frontier.

Key Points

  • Frontier LLM training runs are approaching practical duration limits due to hardware failure rates and cluster reliability at scale.
  • Data availability constrains training length, as high-quality internet text is increasingly exhausted for pretraining.
  • Optimization dynamics mean returns from longer training diminish, limiting the gains from simply extending run duration.
  • This finding has implications for AI scaling trajectories and suggests compute scaling must occur via other dimensions (model size, data quality, new architectures).
  • The analysis uses empirical data on training run lengths and failure rates to bound realistic maximum training durations.

Cited by 1 page

PageTypeQuality
Capability-Alignment Race ModelAnalysis62.0

Cached Content Preview

HTTP 200Fetched Feb 23, 20269 KB
Frontier LLM training runs can’t get much longer | Epoch AI 
 

 
 
 
 
 
 

 

 
 
 
 

 

 

 

 

 
 
 

 

 

 

 
 
 
 

 

 

 

 

 
 
 
 
 
 
 
 
 

 

 
 
 

 
 
 

 
 
 
 
 

 
 

 
 
 
 
 
 
 
 
 

 
 
 

 
 
 
 Latest 
 
 
 
 
 
 
 Publications & Commentary 
 
 
 
 
 
 
 
 Papers & Reports 
 
 
 
 Newsletter 
 
 
 
 Podcast 
 
 
 
 
 
 
 
 
 Data & Resources 
 
 
 
 
 
 
 
 
 Datasets 
 
 Overview 
 
 Benchmarking 
 
 Models 
 
 Frontier Data Centers 
 
 Hardware 
 
 Companies 
 
 Chip Sales 
 
 Polling 
 
 
 
 
 
 
 Resources 
 
 AI Trends & Statistics 
 
 Data Insights 
 
 
 
 
 
 
 
 
 
 
 Projects 
 
 
 
 
 
 
 
 FrontierMath 
 
 
 
 GATE Playground 
 
 
 
 Distributed Training 
 
 
 
 Model Counts 
 
 
 
 
 
 
 
 
 About 
 
 
 
 
 
 
 
 About Us 
 
 
 
 Our Team 
 
 
 
 Careers 
 
 
 
 Consultations 
 
 
 
 Our Funding 
 
 
 
 Donate 
 
 
 
 
 
 
 
 
 Contact 
 
 
 
 
 
 
 
 
 
 
 

 
 
 

 
 

 
 
 
 
 
 
 
 

 
 
 
 
 
 
 

 
 
 
 
 

 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 

 
 
 
 
 

 

 
 
 
 
 
 Search epoch.ai
 
 
 Search 
 

 
 
 
 Enter a query to search for results
 
 
 
 

 
 
 
 Placeholder 
 

 
 
 
 
 

 
 
 

 
 
 
 
 
 
 
 Epoch AI’s work is free to use, distribute, and reproduce provided the source and authors are credited under the Creative Commons Attribution license .

 

 
 
 Cite this work as

 

 
 
 
 Luke Emberson and Yafah Edelman (2025), "Frontier training runs will likely stop getting longer by around 2027". Published online at epoch.ai. Retrieved from: 'https://epoch.ai/data-insights/longest-training-run' [online resource] 
 
 
 
 

 
 BibTeX citation

 

 
 @misc{epoch2025longesttrainingrun,
 title={Frontier training runs will likely stop getting longer by around 2027},
 author={Luke Emberson and Yafah Edelman},
 year={2025},
 url={https://epoch.ai/data-insights/longest-training-run},
 note={Accessed: }
 } 
 

 
 
 

 
 
 
 
 
 

 

 
 
 

 

 
 

 
 Data Insight 
 
 
 
 
 
 

 
 Frontier training runs will likely stop getting longer by around 2027
 
 
 

 Frontier training runs will likely stop getting longer by around 2027

 

 In “ The Longest Training Run ”, we argue that training runs that last too long are outclassed by training runs that start later and benefit from additional hardware and algorithmic improvements . Based on our latest numbers, this suggests that training runs lasting more than 9 months may be inefficient. At the current pace, training runs will reach this size around 2027 (90% CI: Aug 2025 to Sept 2029).

 
 
 
 
 
 
 

 

 
 

 

 

 
 

 

 
 

 

 
 
 
 Enable JavaScript to see an interactive visualization.

 

 
 

 

 
 
 

 
 

 Longer training runs are a significant driver of the rapid growth seen in training compute . If training time stops increasing, training compute growth will slow – unless developers ramp up hardware scaling even faster. This could be achieved by speeding up the build-out of larger clusters, or by spreading training across multiple cluste

... (truncated, 9 KB total)
Resource ID: 9d535d8e91127085 | Stable ID: MzRiNDYyNT