HLE-Diamond - Reasoning: leaderboard
Metric: Accuracy (%, 500 reasoning questions, no tools). Source: lastexam.ai. Saturation forecast: Around April 2027. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-6 (Max) | 81.6 |
| 2 | GPT-6 (High) | 75.6 |
| 3 | Claude Opus 5.5 (Max) | 69.8 |
| 4 | Claude Opus 5.5 (High) | 63.2 |
| 5 | Claude Fable 5.1 (Max) | 62.4 |
| 6 | Claude Fable 5.1 (High) | 62 |
| 7 | GPT-6 Sol (Max) | 58.4 |
| 8 | GPT-5.6 Sol (Max) | 54.4 |
| 9 | Claude Opus 5 (Max) | 47.8 |
| 10 | Claude Opus 5 (High) | 47 |
| 11 | Muse Spark 1.3 (Max) | 45 |
| 12 | GPT-6 Sol (High) | 44.2 |
| 13 | GPT-5.6 Sol (High) | 38.8 |
| 14 | Gemini 3.8 Flash (High) | 38.6 |
| 15 | Grok 4.7 (xHigh) | 34.6 |
Interactive version: theaggregate.ai/benchmark?slug=hle-diamond-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-26.