o1: capability breakdown
99.7 / 220
measured 13 of 22 rungs · effort row: default (the median of its published effort settings)
This page shows where the number comes from. Each rung is one public benchmark, worth up to 10 points times the model's chance-corrected score on it; the score is the sum of all 22 rungs. How the capability ladder works explains the rungs, the fit and its limits.
| Rung (easiest → hardest) | Points / 10 | How | Source date |
|---|---|---|---|
| HumanEval | 9.54 | measured | 2024-10-05 |
| MMLU | 8.77 | measured | 2024-10-05 |
| MATH-500 | 9.47 | measured | 2025-04-14 |
| MMLU-Pro | 8.46 | fitted | |
| AIME 2024 | 9.55 | fitted | |
| AIME 2025 | 8.23 | fitted | |
| LiveCodeBench | 6.79 | measured | 2026-09-05 |
| GPQA Diamond | 5.59 | measured | 2024-10-05 |
| IFBench | 7.03 | measured | 2026-04-16 |
| τ²-Telecom | 6.26 | measured | 2026-04-16 |
| AA-LCR | 5.93 | measured | 2026-04-16 |
| Terminal-Bench 2.1 | 2.89 | fitted | |
| SciCode | 3.58 | measured | 2026-04-16 |
| Terminal-Bench Hard | 1.29 | measured | 2026-04-16 |
| AutomationBench | 0.54 | fitted | |
| AA-Briefcase (rubric pass) | 0.63 | fitted | |
| HLE | 0.30 | measured | 2026-04-16 |
| MLCR | 0.12 | fitted | |
| τ³-Banking | 0.85 | fitted | |
| Omniscience acc. | 3.47 | measured | 2026-04-16 |
| CritPt | 0.03 | measured | 2026-04-16 |
| GDP.pdf | 0.38 | fitted | |
| Total | 99.70 | = 99.7 / 220 | |
Measured rows are the model's own published result on that benchmark. fitted rows are an item-response estimate: what a model with this ability would be expected to score on a benchmark it has no published result for (see the methodology).
Per-benchmark results are Artificial Analysis's (public model pages); fitted values are this site's estimates. Permission to republish requested 2026-09-05.