NuclearQAv2: leaderboard
NuclearQAv2 (2026) is a nuclear engineering question-answering benchmark of about 1,240 questions in three formats: yes-or-no factual questions, numeric calculations and short verbal concept answers. The score is aggregated accuracy in percent over all questions, averaged over three runs; higher is better.
Metric: Aggregated accuracy (%) over all 1,239 questions (750 boolean, 206 numeric, 283 verbal); mean of three runs; answers requested without reasoning text. Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-OSS-120B | 79.23 | #700 |
| 2 | GPT-5.4 | 78.91 | #123 |
| 3 | GPT-5.2 | 78.42 | #176 |
| 4 | NVIDIA-Nemotron-3-Super-120B-A12B-FP8 | 78.4 | #469 |
| 5 | GPT-4o | 72.45 | #741 |
| 6 | Llama 3 70B Instruct | 70.92 | #1676 |
| 7 | Llama 3 8B Instruct | 62.36 | #4612 |
| 8 | Nova Pro | 60.37 | #1555 |
| 9 | Mistral 7B Instruct | 35.48 | #7468 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=nuclearqav2 · How It Works · Data refreshed daily, snapshot 2026-11-13.