HLE-Diamond (With Tools): leaderboard
HLE-Diamond's 1,000 questions answered with web search and code execution at reasoning high, under CAIS's recommended tools settings. Compare it with other tool-enabled runs, not the closed-book board.
Metric: Accuracy (%, web + code tools). Source: lastexam.ai. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (High) | 82.9 | #6 (GPT-6 Astra) |
| 2 | Claude Opus 5.5 (High) | 73.9 | #5 (Claude Opus 5.5) |
| 3 | Claude Fable 5.1 (High) | 72.3 | #14 (Claude Fable 5.1) |
| 4 | GPT-6.1 Sol (High) | 69.1 | #10 (GPT-6.1 Sol) |
| 5 | Claude Opus 5 (High) | 68.9 | #20 (Claude Opus 5) |
| 6 | GPT-6 Sol (High) | 65 | #24 (GPT-6 Sol) |
| 7 | GPT-5.6 Sol (High) | 56.6 | #29 (GPT-5.6 Sol) |
| 8 | Muse Spark 1.3 (High) | 55.4 | #42 (Muse Spark 1.3) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=hle-diamond-with-tools · How It Works · Data refreshed daily, snapshot 2026-11-12.