87.8% on FrontierMath T4.
Research-level problems set by working mathematicians, and contest problems below them.
Clears the bar in
2026
80% interval 2026–2026
20252028
90% of FrontierMath Tier 4: nine in ten research-level problems, each a specialist's day of work is the bar this date is solved against.
⚠ FrontierMath Tier 4 accuracy needs 37 of 41 items and the latest reading has 36: 1 more. One run of the same model moves by about ±2 items on a test this size, so this domain is at its bar within measurement noise, and the year below is a forecast of that noise rather than of new capability.
The measurement
Each dot is one model's FrontierMath Tier 4 accuracy, plotted on its release date. Nothing on this chart is a forecast.
- Doubling
- ~2.0 mo
- 62 days · all measurements
- Fit
- r² 0.96
- 8 of 8 measurements fitted
- Measured span
- -0.0% 87.8%
- Jan 2025 → Jun 2026
How to read this chart
The axis is a log-odds scale, the same one the trend is fit on. Equal distance means an equal cut in the remaining error, so 50% to 90% is about the same step as 90% to 99%, which is why the labels crowd near the top. A percentage axis would flatten near 100% whether or not capability flattened. Very close to the ceiling the steps shorten, because a 41-question test cannot resolve past its own last item.
Our fit over all measurements doubles every 62 days. Epoch AI · Benchmarking Hub publish the scores and the release dates. The trend line through them is ours, and it is the only thing on this chart that is.
Measurements
8 modelsEvery point on the chart above, named. The date is the model's release, which is what the trend is regressed on.
| Model | Released | Position on the measured range | FrontierMath T4 |
|---|---|---|---|
| Claude Fable 5 | Jun 2026 | 87.8% | |
| GPT-5.5 Pro | Apr 2026 | 78.0% | |
| GPT-5.4 | Mar 2026 | 49.0% | |
| GPT-5.4 Pro | Mar 2026 | 58.5% | |
| GPT-5.2 Pro | Dec 2025 | 46.0% | |
| GPT-5 | Aug 2025 | 22.0% | |
| o4-mini | Apr 2025 | 4.9% | |
| o3-mini | Jan 2025 | 0.0% |
Questions
FrontierMath is answered by the same frontier language models Epoch's largest-training-run series describes, so the systems measured and the systems setting the compute frontier are the same class. Contest sets such as Mock AIME are solved outright by frontier models, so they have no signal left to fit, which is why this domain is measured on the research-level tier.
The whole Tier 4 series is post-2025, because it did not exist earlier, so there is no earlier era to split off. Every SOTA measurement is fit.
No. Each domain is measured on its own scale - a task length in hours, an accuracy on a fixed test set - and every one of them scores a model on a bounded, pre-specified set of tasks. That is a measure of DEPTH on that set, not of breadth of competence. Depth is not breadth: a model can clear the bar on FrontierMath T4 and still fail at things a child does.