ScienceArena - APhO 2025: leaderboard
The Asian Physics Olympiad 2025 theory exam in ScienceArena (Zhao et al., 2026), solved one subpart at a time from text renderings of the questions. Points out of 30, awarded by an LLM judge applying the official process-credit rubric, calibrated against medalist grading; higher is better.
Metric: Score (out of 30): points awarded on the Asian Physics Olympiad 2025 theory problems, text-only question renderings; interleaved solving, one subpart at a time with earlier turns in context; open-ended physics and chemistry turns graded against the official solution and process-credit rubric by the authors' medalist-calibrated LLM-judge pipeline. Source: arxiv.org. Saturation forecast: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 28.5 | #94 |
| 2 | GLM-5.1 | 28.4 | #179 |
| 3 | Qwen 3.7 Max | 28.1 | #145 |
| 4 | Gemini 3.5 Flash | 27.91 | #101 |
| 5 | Qwen 3.6 Plus | 27.56 | #222 |
| 6 | Gemini 3 Flash | 26.75 | #164 |
| 7 | DeepSeek V4 Pro | 26.01 | #188 |
| 8 | Claude Opus 4.7 | 24.19 | #79 |
| 9 | MiniMax-M2.7 | 22.56 | #518 |
| 10 | GPT-5.4 | 22.01 | #123 |
| 11 | Seed 2.0 Pro | 21.72 | #211 |
| 12 | GPT-5.5 | 21.5 | #49 |
| 13 | MiMo-V2.5-Pro | 21.33 | #279 |
| 14 | DeepSeek V4 Flash | 20.44 | #258 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=sciencearena-apho-2025 · How It Works · Data refreshed daily, snapshot 2026-11-13.