Nemotron 3 Nano 30B A3B (Non-reasoning): benchmark results
Provider: NVIDIA. Released 2025-12-15. Access: Open.
Unified ELO 1414 ± 17, rank #1294 of 2055 rated models, from 80 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| UnpredictaBench - Code | 18.49 | KS@100 (%): share of the 238 Code problems (distributions in | 89.5 |
| UnpredictaBench - RealWorld | 33.33 | KS@100 (%): share of the 30 RealWorld problems (natural-lang | 86.8 |
| UnpredictaBench - Shuffling | 31.58 | KS@100 (%): share of the 19 Shuffling problems (random order | 78.9 |
| UnpredictaBench - Text | 20.63 | KS@100 (%): share of the 160 Text problems (named or describ | 68.4 |
| AA Terminal-Bench Hard | 12.12 | Accuracy (%) | 48 |
| AA LiveCodeBench | 35.98 | Pass@1 (%) | 43.9 |
| AA Omniscience - Software Engineering (SWE) - Python | 14 | Accuracy (%) | 34.5 |
| AA IFBench | 37.48 | Accuracy (%) | 31.6 |
| Tau2-Bench Telecom | 25.44 | Success Rate (%) | 30.3 |
| AA Omniscience - Software Engineering (SWE) - HTML | 22.45 | Accuracy (%) | 29.4 |
| AA TAU-2 Bench | 25.44 | Accuracy (%) | 29.4 |
| AA Omniscience - Software Engineering (SWE) - TypeScript | 11.11 | Accuracy (%) | 29.2 |
Interactive version: theaggregate.ai/model?slug=nemotron-3-nano-30b-a3b-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-29.