According to our Epoch Capabilities Index, frontier model improvement nearly doubled in pace, from ~8 points/year prior to April 2024, to ~15 points/year thereafter.
This acceleration roughly coincides with the rise of reasoning models and increased focus on reinforcement learning at frontier labs. The METR Time Horizon benchmark shows a similar pattern, with a 40% acceleration in October 2024.
We extended our ECI data as far back as December 2021 to better track changes in capabilities. Note that this earlier data has high uncertainty, and caution should be used when interpreting ECI estimates from before 2023.
Read more about Jaeho and Yafah’s analysis, and explore the interactive plot on our website!
Or, check out the ECI on our benchmarking datahub.