Apriel-v1.5-15B-Thinker — benchmark results
Provider: ServiceNow. Released 2025-06-01. Access: Open.
Unified ELO 1560 ± 46, rank #614 of 1841 rated models, from 54 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA AIME 2025 | 87.5 | Accuracy (%) | 84.4 |
| AA LiveCodeBench | 72.8 | Pass@1 (%) | 83.6 |
| CritPt | 1.1 | Accuracy (self-reported) | 77.4 |
| AA IFBench | 61.7 | Accuracy (%) | 72.8 |
| AA CritPt | 1.14 | Accuracy (%) | 71.4 |
| AA Humanity's Last Exam | 12.05 | Accuracy (%) | 67.2 |
| Artificial Analysis Intelligence Index | 21.23 | Intelligence Index | 61.6 |
| AA TAU-2 Bench | 68.42 | Accuracy (%) | 61.5 |
| AA MMLU-Pro | 77.34 | Accuracy (%) | 56.2 |
| AA GPQA Diamond | 71.31 | Accuracy (%) | 54.8 |
| AA Omniscience - Science, Engineering & Mathematics | 26.7 | Accuracy (%) | 53.4 |
| Epoch AI - Scicode | 34.84 | Score | 51.8 |
Interactive version: theaggregate.ai/model?slug=apriel-v1-5-15b-thinker · How It Works · Data refreshed daily, snapshot 2026-07-25.