reasoning benchmark
BIG-Bench Extra Hard
BIG-Bench Extra Hard (BBEH) is a challenging benchmark that replaces each task in BIG-Bench Hard with a novel task that probes similar reasoning capabilities but exhibits significantly increased difficulty. The benchmark contains 23 tasks testing diverse reasoning skills including many-hop reasoning, causal understanding, spatial reasoning, temporal arithmetic, geometric reasoning, linguistic reasoning, logic puzzles, and humor understanding. Designed to address saturation on existing benchmarks where state-of-the-art models achieve near-perfect scores, BBEH shows substantial room for improvement with best models achieving only 9.8-44.8% average accuracy.
Updated Aug 7, 2026
Published models11
Registry coverage11
MetricScore
EvidenceB
BIG-Bench Extra Hard leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | GO | 74.4% | 100.0% | 11 | C | |
| 2 | GO | 64.8% | 90.0% | 11 | C | |
| 3 | GO | 53.0% | 80.0% | 11 | C | |
| 4 | GO | 47.6% | 70.0% | 11 | C | |
| 5 | GO | 33.1% | 60.0% | 11 | C | |
| 6 | GO | 21.9% | 50.0% | 11 | C | |
| 7 | GO | 19.3% | 40.0% | 11 | C | |
| 8 | GO | 16.3% | 30.0% | 11 | C | |
| 9 | GO | 15.0% | 20.0% | 11 | C | |
| 10 | GO | 11.0% | 10.0% | 11 | C | |
| 11 | GO | 7.2% | 0.0% | 11 | C |
Score distribution
Top published rows on the benchmark's original scale.
BIG-Bench Extra Hard highlights
The top published results on this benchmark's own scale.
What is BIG-Bench Extra Hard?
Definition and scoring fields from the benchmark registry.
BIG-Bench Extra Hard (BBEH) is a challenging benchmark that replaces each task in BIG-Bench Hard with a novel task that probes similar reasoning capabilities but exhibits significantly increased difficulty. The benchmark contains 23 tasks testing diverse reasoning skills including many-hop reasoning, causal understanding, spatial reasoning, temporal arithmetic, geometric reasoning, linguistic reasoning, logic puzzles, and humor understanding. Designed to address saturation on existing benchmarks where state-of-the-art models achieve near-perfect scores, BBEH shows substantial room for improvement with best models achieving only 9.8-44.8% average accuracy.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- BIG-Bench Extra Hard
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- big-bench-extra-hard|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about BIG-Bench Extra Hard.
Which model scores highest on BIG-Bench Extra Hard?
Gemma 4 31B is currently ranked first with 74.4%.
What does BIG-Bench Extra Hard measure?
BIG-Bench Extra Hard (BBEH) is a challenging benchmark that replaces each task in BIG-Bench Hard with a novel task that probes similar reasoning capabilities but exhibits significantly increased difficulty. The benchmark contains 23 tasks testing diverse reasoning skills including many-hop reasoning, causal understanding, spatial reasoning, temporal arithmetic, geometric reasoning, linguistic reasoning, logic puzzles, and humor understanding. Designed to address saturation on existing benchmarks where state-of-the-art models achieve near-perfect scores, BBEH shows substantial room for improvement with best models achieving only 9.8-44.8% average accuracy.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
11 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is marked as eligible for the current LLMBoard capability methodology.