Epoch AI
Subscribe
Sign in
Home
Gradient Updates
Data Insights
Papers & Reports
Epoch After Hours
The Epoch Brief
Updates
Archive
About
Latest
Top
Discussions
What do “economic value” benchmarks tell us?
These benchmarks track a wide range of digital work. Progress will correlate with economic utility, but tasks are too self-contained to indicate full…
18 hrs ago
•
Florian Brand
and
Greg Burnham
9
1
How close is AI to taking my job?
Beyond benchmarks as leading indicators for task automation
Feb 7
•
Anson Ho
42
8
6
Watch or listen — Do AI models cover their costs?
Fri Feb 6th recording now available
Feb 7
•
Epoch AI & various writers
Join this Friday — Do AI models cover their costs?
Fri Feb 6th @9am PT (noon ET / 5pm GMT / 6pm CET)
Feb 5
•
Epoch AI & various writers
4
1
Kimi K2.5 has highest ECI score among open weight models
Its score of 147 is about on par with o3, Grok 4, and Sonnet 4.5. It still lags the overall frontier.
Feb 4
1
1
The Epoch AI Brief - January 2026
AI Chip Sales data explorer launch! FrontierMath: Open Problems benchmark, global AI power capacity, OpenAI profitability, how well last year's…
Feb 2
•
Epoch AI & various writers
16
2
January 2026
AI math capabilities could be jagged for a long time – Daniel Litt
How does math research change when the cost of trying your first dumb idea goes to zero?
Jan 29
•
Anson Ho
and
Greg Burnham
13
2
Can AI companies become profitable?
Lessons from GPT-5’s economics
Jan 28
•
Jaime Sevilla
,
Hannah Petrovic
, and
Anson Ho
44
7
6
Introducing FrontierMath: Open Problems
Benchmarking AI on unsolved research problems that have eluded mathematicians
Jan 27
18
1
Epoch's Latest AI Trends
We’ve added new trends & figures categories to our Trends page!
Jan 23
•
Luke Emberson
13
1
New record on FrontierMath Tier 4
GPT-5.2 Pro (manual run) scored 31%, a substantial jump over the previous high score of 19%.
Jan 23
15
1
1
Polling on AI Usage
We collaborated with the polling firm Blue Rose Research to survey 5,660 Americans about their AI usage habits in late 2025.
Jan 21
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts
