Nemotron 3 Nano Omni 30B-A3B (Reasoning): benchmark results
Nemotron 3 Nano Omni 30B-A3B (Reasoning) supports video, audio, image and text input with text output, 256K context, and approximately 3B active parameters per generated token. The publisher describes a 31B hybrid MoE backbone; encoder processing adds modality-dependent work. Provider: NVIDIA. Released 2026-04-28. Access: Open.
Unified ELO 1493 ± 1, rank #1531 of 3078 rated models, from 54 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA IFBench | 63.2 | Accuracy (%) | 74.2 |
| AA Omniscience - Software Engineering (SWE) - Kotlin | 20 | Accuracy (%) | 63 |
| AA Omniscience - Software Engineering (SWE) - Rust | 50 | Accuracy (%) | 59 |
| AA Omniscience - Software Engineering (SWE) - Swift | 32 | Accuracy (%) | 51.2 |
| Tau2-Bench Telecom | 45.32 | Success Rate (%) | 50.6 |
| AA Omniscience - Software Engineering (SWE) - Java | 15 | Accuracy (%) | 49.8 |
| AA TAU-2 Bench | 45.32 | Accuracy (%) | 49.1 |
| AA Omniscience - Software Engineering (SWE) - Julia | 8 | Accuracy (%) | 46.1 |
| Artificial Analysis Intelligence Index | 10.25 | Intelligence Index | 45.9 |
| AA Terminal-Bench Hard | 8.33 | Accuracy (%) | 41.8 |
| AA-Omniscience Hallucination Rate | 85.65 | Hallucination Rate (%) | 41.4 |
| AA-Omniscience Index - Software Engineering (SWE) - Rust | 4 | Omniscience Index | 40 |
Interactive version: theaggregate.ai/model?slug=nemotron-3-nano-omni-30b-a3b-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-19.