Organisation
Epoch AI
Nonprofit research institute that measures and forecasts the trajectory of AI — compute, data, hardware and investment — and built the FrontierMath benchmark.
Epoch AI is a research organisation that tracks the numbers behind AI progress — training compute, data, hardware, investment and how quickly capabilities are advancing — and its datasets are widely cited by researchers, journalists and policymakers. In late 2024 it built FrontierMath, a benchmark of original, expert-level mathematics problems designed to resist the pattern-matching that had let models saturate earlier tests; leading systems initially solved under 2%. The benchmark also drew Epoch into controversy: in early 2025 it emerged that OpenAI had funded FrontierMath and held access to its problems and solutions, a relationship Epoch had not disclosed to the mathematicians who built it, which the organisation acknowledged as a mistake on transparency. It continues to publish influential analysis on the trajectory and economics of frontier AI.
- Category
- Evaluation & forecasting
- Founded
- 2022
- Key people
- Jaime Sevilla
Appears alongside
Featured in threads
Tracks
- Benchmarks & progress 8
- Compute & infrastructure 3
- Open weights & ecosystem 2
- Ideas & essays 2
- Government & policy 1
- Money & business 1
Epoch AI expands FrontierMath to 50 unsolved research problems
Unlike FrontierMath's original tiers, these problems have no known solution at all; three of the fifty have been solved by AI, including one by GPT-5.6 Sol.
Benchmarks & progress
Epoch AI: open models lag closed frontier by four months
The gap, measured on Epoch's Capabilities Index, has held at roughly eight index points since January 2026 — comparable to the distance between GPT-5 and GPT-5.5.
Benchmarks & progress · Open weights & ecosystem
Epoch AI finds fatal errors in about a third of FrontierMath problems
Most flagged errors were simple mistakes in the published answer key — off-by-one slips and flipped signs — rather than genuinely ambiguous problems, Epoch said.
Benchmarks & progress
Epoch AI estimates scale of smuggled-chip compute in China
A 90% confidence interval ran from 290,000 to 1.6 million smuggled H100-equivalents, with a median estimate — about a third of China's compute — built from unproven indictments and resale-market tracking.
Compute & infrastructure · Government & policy
Epoch AI: Anthropic revenue closing in on OpenAI's
Epoch AI put Anthropic's annualised revenue growth at roughly 10x a year since reaching $1 billion, against about 3.4x for OpenAI, projecting a possible crossover around mid-2026.
Money & business
Epoch AI reports AI capabilities progress has sped up
A two-segment regression across 149 models found capability gains almost doubled in pace after April 2024, a break Epoch linked to the rise of reasoning models.
Benchmarks & progress
Epoch AI reports open models trail closed models by about 3.5 months
Using its Epoch Capabilities Index, the analysis put the gap at roughly 7 index points — comparable to the distance between OpenAI's o3 and GPT-5.
Benchmarks & progress · Open weights & ecosystem
Epoch AI reports GPT-5's FrontierMath performance
Running its own scaffold rather than OpenAI's, Epoch scored GPT-5 at 24.8% on FrontierMath's main tiers and 8.3% on the hardest tier, a new high for the benchmark.
Benchmarks & progress
Epoch AI's undisclosed OpenAI funding of FrontierMath draws criticism
Epoch AI acknowledged OpenAI funded and had privileged access to FrontierMath's problems and solutions, and had not told contributing mathematicians before the benchmark featured in o3's launch.
Benchmarks & progress
Epoch AI launches FrontierMath
Built with over 60 mathematicians including Fields medallists as reviewers, the benchmark held leading models under 2% accuracy even with extended reasoning time and code tools.
Benchmarks & progress
Epoch AI publishes 'Training compute of frontier AI models grows by 4-5x per year'
The estimate drew on 333 compute figures for notable models since 2010 — roughly triple Epoch's 2022 dataset — and flagged an unexplained slowdown around 2018.
Ideas & essays · Compute & infrastructure
Epoch AI publishes 'Will We Run Out of Data?'
Modelled dataset growth against the finite stock of public text and projected the two curves would cross between 2026 and 2032, sooner if models were overtrained.
Ideas & essays · Compute & infrastructure
Also mentioned in 1 entry
Referenced in passing — Epoch AI isn't the main subject of these.