OpenCompass Research - HLE: leaderboard
OpenCompass research-table score on Humanity's Last Exam, covering expert-level questions across broad academic domains.
Metric: Score (%). Source: rank.opencompass.org.cn. Saturation forecast: Around March 2027. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 38 | #66 |
| 2 | Kimi K2.5 | 28.6 | #144 |
| 3 | DeepSeek V3.2 Speciale | 28.6 | #128 |
| 4 | GPT-5 (2025-08-07) (High) | 28.5 | |
| 5 | GLM-5 | 28.1 | #141 |
| 6 | Qwen 3.5 397B A17B | 27.5 | #146 |
| 7 | GLM-4.7 | 25.4 | #192 |
| 8 | DeepSeek V3.2 | 23.2 | #205 |
| 9 | MiniMax-M2.5 | 22.2 | #303 |
| 10 | Step 3.5 Flash | 21.6 | #270 |
| 11 | GPT-5 Mini (2025-08-07) (High) | 21.5 | #172 (GPT-5 Mini (2025-08-07)) |
| 12 | O3 (2025-04-16) | 21.3 | #121 |
| 13 | Kimi K2 (Thinking) | 21.3 | #243 (Kimi K2) |
| 14 | Gemini 2.5 Pro | 21.1 | #152 |
| 15 | MiMo-V2-Flash | 20.5 | #372 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=opencompass-research-hle · How It Works · Data refreshed daily, snapshot 2026-10-20.