BBEH: leaderboard
BIG-Bench Extra Hard: 27 tasks from Google DeepMind that push frontier models beyond standard BIG-Bench Hard.
Metric: Harmonic Mean (%). Source: github.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 Mini (High) | 44.8 | #272 (O3 Mini) |
| 2 | Gemini 2.0 Flash | 9.8 | #342 |
| 3 | Gemini 2.0 Flash Lite | 8 | #450 |
| 4 | DeepSeek R1 | 6.8 | #251 |
| 5 | GPT-4o | 6 | #343 |
| 6 | Gemma 3 27B | 4.9 | #607 |
| 7 | Gemma 3 12B | 4.5 | #671 |
| 8 | Gemma 2 27B (IT) | 4 | #645 |
| 9 | Llama 3.1 8B Instruct | 3.6 | #1035 |
| 10 | Gemma 3 4B | 3.4 | #1089 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=bbeh · How It Works · Data refreshed daily, snapshot 2026-10-20.