language benchmark
BIG-Bench Hard (BBH) is a subset of 23 challenging BIG-Bench tasks with 6,511 evaluation examples covering multiple forms of multi-step reasoning.
Updated Sep 5, 2026
Higher score ranks better on this benchmark.
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAN | Score93.10% | Percentile100.00% | Participants21 | EvidenceC | Evaluated |
| Rank02 | ModelAN | Score93.10% | Percentile95.00% | Participants21 | EvidenceC | Evaluated |
| Rank03 | ModelGO | Score89.20% | Percentile90.00% | Participants21 | EvidenceC | Evaluated |
| Rank04 | ModelGO | Score87.60% | Percentile85.00% | Participants21 | EvidenceC | Evaluated |
| Rank05 | ModelAN | Score86.80% | Percentile80.00% | Participants21 | EvidenceC | Evaluated |
| Rank06 | ModelGO | Score85.70% | Percentile75.00% | Participants21 | EvidenceC | Evaluated |
| Rank07 | ModelGO | Score85.50% | Percentile70.00% | Participants21 | EvidenceC | Evaluated |
| Rank08 | ModelAN | Score82.90% | Percentile65.00% | Participants21 | EvidenceC | Evaluated |
| Rank09 | ModelMI | Score79.10% | Percentile60.00% | Participants21 | EvidenceC | Evaluated |
| Rank10 | ModelAN | Score73.70% | Percentile55.00% | Participants21 | EvidenceC | Evaluated |
| Rank11 | ModelGO | Score72.20% | Percentile50.00% | Participants21 | EvidenceC | Evaluated |
| Rank12 | ModelMI | Score70.40% | Percentile45.00% | Participants21 | EvidenceC | Evaluated |
| Rank13 | ModelIB | Score69.13% | Percentile40.00% | Participants21 | EvidenceC | Evaluated |
| Rank14 | ModelIB | Score69.13% | Percentile35.00% | Participants21 | EvidenceC | Evaluated |
| Rank15 | ModelMI | Score69.00% | Percentile30.00% | Participants21 | EvidenceC | Evaluated |
| Rank16 | ModelIB | Score55.70% | Percentile25.00% | Participants21 | EvidenceC | Evaluated |
| Rank17 | ModelGO | Score52.90% | Percentile20.00% | Participants21 | EvidenceC | Evaluated |
| Rank18 | ModelGO | Score52.90% | Percentile15.00% | Participants21 | EvidenceC | Evaluated |
| Rank19 | ModelGO | Score44.30% | Percentile10.00% | Participants21 | EvidenceC | Evaluated |
| Rank20 | ModelGO | Score44.30% | Percentile5.00% | Participants21 | EvidenceC | Evaluated |
| Rank21 | ModelGO | Score39.10% | Percentile0.00% | Participants21 | EvidenceC | Evaluated |
The leading models and scores on this benchmark.
A closer view of the leading scores on this benchmark.
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What BIG-Bench Hard measures and how its scores work.
BIG-Bench Hard (BBH) is a language benchmark consisting of 23 BIG-Bench tasks selected because prior language model evaluations did not outperform average human-rater performance.
It measures Score as a ratio across arithmetic, logical, geometric, and temporal reasoning, language understanding, causal judgment, object counting, navigation, pattern recognition, and complex problem solving.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
Common questions about BIG-Bench Hard.
Claude 3.5 Sonnet is currently ranked first with 93.10%.
The current leaders are Claude 3.5 Sonnet (93.10%), Claude 3.5 Sonnet (93.10%), and Gemini 1.5 Pro (89.20%).
Phi 4 Mini has the lowest matched official input price at $0.08 input / $0.30 output per 1M tokens.
The fastest matched records are Gemini 1.5 Flash (150.00 tok/s via Google), Claude 3.5 Sonnet (100.00 tok/s via Anthropic), and Claude 3 Opus (100.00 tok/s via Anthropic).
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.
It measures Score as a ratio across arithmetic, logical, geometric, and temporal reasoning, language understanding, causal judgment, object counting, navigation, pattern recognition, and complex problem solving.
Yes. Higher values rank better for this benchmark.
21 model results are currently shown.
No. This benchmark is shown for reference but does not contribute to the overall score.
Ranking basisThis big-bench hard AI model leaderboard uses descending score in the benchmark original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
Selection summary
Claude 3.5 Sonnet currently leads BIG-Bench Hard with 93.10%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price, and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.