reasoning benchmark
WideSearch
WideSearch is an agentic search benchmark that evaluates models' ability to perform broad, parallel search operations across multiple sources. It tests wide-coverage information retrieval and synthesis capabilities.
Updated Aug 7, 2026
Published models9
Registry coverage9
MetricScore
EvidenceB
WideSearch leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | MA | 80.8% | 100.0% | 9 | C | |
| 2 | MA | 79.0% | 87.5% | 9 | C | |
| 3 | TE | 76.4% | 75.0% | 9 | C | |
| 4 | AC | 74.3% | 62.5% | 9 | C | |
| 5 | AC | 74.0% | 50.0% | 9 | C | |
| 6 | AC | 61.1% | 37.5% | 9 | C | |
| 7 | AC | 60.5% | 25.0% | 9 | C | |
| 8 | AC | 60.1% | 12.5% | 9 | C | |
| 9 | AC | 57.1% | 0.0% | 9 | C |
Score distribution
Top published rows on the benchmark's original scale.
WideSearch highlights
The top published results on this benchmark's own scale.
What is WideSearch?
Definition and scoring fields from the benchmark registry.
WideSearch is an agentic search benchmark that evaluates models' ability to perform broad, parallel search operations across multiple sources. It tests wide-coverage information retrieval and synthesis capabilities.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- WideSearch
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- widesearch|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about WideSearch.
Which model scores highest on WideSearch?
Kimi K2.6 is currently ranked first with 80.8%.
What does WideSearch measure?
WideSearch is an agentic search benchmark that evaluates models' ability to perform broad, parallel search operations across multiple sources. It tests wide-coverage information retrieval and synthesis capabilities.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
9 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.