DeepSeek-R1 is an open-weights model from DeepSeek, released January 20, 2025. It has an ECI score of 139 and ranks 120th of 253 tracked models. At the time of its release, it ranked 2nd of 105 tracked models.
| Domain | Benchmark | ||||
|---|---|---|---|---|---|
ECI scoreBest score 167 | 139 | 132 | 142 | 135 | |
| Ranking | #120 | #147 | #102 | #140 | |
| Conceptual reasoning | SimpleBenchBest score 82% | ||||
ARC-AGI-1Best score 99% | |||||
| Mathematics | MATH Level 5Best score 98% | ||||
OTIS Mock AIME 2024-2025Best score 100% | |||||
| Science | GPQA DiamondBest score 96% | ||||
| Software engineering | METR Time HorizonsBest score 17 hr 25 min | 26 min 56 sec | 18 min 28 sec | 38 min 50 sec | — |
Aider PolyglotBest score 88% | |||||
WeirdML (v2)Best score 94% | |||||
| Other | LiveBenchBest score 82% |