XpertBench - Engineering and Applied Sciences: leaderboard
Metric: Weighted rubric score (%) on the XpertBench-Gold subset (245 expert-written open-ended professional tasks, 15-40 weighted binary checkpoints each); ShotJudge: Gemini 2.5 Pro judges every checkpoint with an expert-graded GPT-5 response as one-shot exemplar; engineering and applied sciences tasks; higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 (Thinking) | 49.58 |
| 2 | GPT-5 (High) | 48.2 |
| 3 | GPT-5.2 (High) | 46.13 |
| 4 | GPT-5.4 (High) | 42.84 |
| 5 | Gemini 3.1 Pro (Preview) | 37.3 |
| 6 | Kimi K2 (Thinking) | 37.1 |
| 7 | Gemini 3 Pro | 36.34 |
| 8 | DeepSeek V3.1 Terminus (Thinking) | 32.2 |
| 9 | Gemini 3 Flash | 31.56 |
| 10 | Doubao-Seed-1.6 (Thinking) | 29.5 |
| 11 | Claude Sonnet 4.5 (Thinking) | 27.5 |
| 12 | GLM-4.6 | 27.4 |
| 13 | Gemini 2.5 Pro | 25.5 |
Interactive version: theaggregate.ai/benchmark?slug=xpertbench-engineering-and-applied-sciences · How It Works · Data refreshed daily, snapshot 2026-10-07.