NVIDIA Nemotron 3 Super (Thinking) — benchmark results
Provider: NVIDIA. Released 2025-12-15. Access: Open.
Unified ELO 1649 ± 41, rank #384 of 1841 rated models, from 37 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA IFBench | 71.5 | Accuracy (%) | 87.8 |
| AA CritPt | 3.14 | Accuracy (%) | 81 |
| AA Omniscience - Software Engineering (SWE) - Julia | 32 | Accuracy (%) | 79.8 |
| AA Humanity's Last Exam | 19.18 | Accuracy (%) | 78.6 |
| AA Omniscience - Science, Engineering & Mathematics | 34.7 | Accuracy (%) | 74.2 |
| AA GPQA Diamond | 80 | Accuracy (%) | 72.6 |
| AA Omniscience - Software Engineering (SWE) - TypeScript | 37.78 | Accuracy (%) | 72.1 |
| AA Long Context Reasoning | 60 | Accuracy (%) | 71.5 |
| AA Omniscience - Software Engineering (SWE) - HTML | 44 | Accuracy (%) | 71.2 |
| AA Omniscience - Software Engineering (SWE) - Rust | 60 | Accuracy (%) | 71.1 |
| AA Terminal-Bench Hard | 28.79 | Accuracy (%) | 70.8 |
| AA Omniscience - Software Engineering (SWE) - Swift | 48 | Accuracy (%) | 68.7 |
Interactive version: theaggregate.ai/model?slug=nvidia-nemotron-3-super-thinking · How It Works · Data refreshed daily, snapshot 2026-07-25.