reasoning benchmark
Wild Bench
WildBench is an automated evaluation framework that benchmarks large language models using 1,024 challenging, real-world tasks selected from over one million human-chatbot conversation logs. It introduces two evaluation metrics (WB-Reward and WB-Score) that achieve high correlation with human preferences and uses task-specific checklists for systematic evaluation.
Updated Aug 7, 2026
Published models8
Registry coverage8
MetricScore
EvidenceB
Wild Bench leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | MA | 68.5% | 100.0% | 8 | C | |
| 2 | MA | 68.5% | 85.7% | 8 | C | |
| 3 | MA | 66.8% | 71.4% | 8 | C | |
| 4 | MA | 65.3% | 57.1% | 8 | C | |
| 5 | MA | 56.8% | 42.9% | 8 | C | |
| 6 | MA | 52.2% | 28.6% | 8 | C | |
| 7 | AL | 48.5% | 14.3% | 8 | C | |
| 8 | AL | 42.4% | 0.0% | 8 | C |
Score distribution
Top published rows on the benchmark's original scale.
Wild Bench highlights
The top published results on this benchmark's own scale.
What is Wild Bench?
Definition and scoring fields from the benchmark registry.
WildBench is an automated evaluation framework that benchmarks large language models using 1,024 challenging, real-world tasks selected from over one million human-chatbot conversation logs. It introduces two evaluation metrics (WB-Reward and WB-Score) that achieve high correlation with human preferences and uses task-specific checklists for systematic evaluation.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- Wild Bench
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- wild-bench|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about Wild Bench.
Which model scores highest on Wild Bench?
MiniStral 3 (14B Instruct 2512) is currently ranked first with 68.5%.
What does Wild Bench measure?
WildBench is an automated evaluation framework that benchmarks large language models using 1,024 challenging, real-world tasks selected from over one million human-chatbot conversation logs. It introduces two evaluation metrics (WB-Reward and WB-Score) that achieve high correlation with human preferences and uses task-specific checklists for systematic evaluation.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
8 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.