DeepResearch Bench
Benchmarks
TaskDatasetSOTA resultTrendResults
DeepResearch Bench
52.84Comprehensiveness
89
DeepResearch Bench official 100-task-subset 1.0
0.508RACE Overall
24
DeepResearch Bench
48.27Overall Score
20
DeepResearch Bench 2025 (test)
49.5Comprehensiveness
16
DeepResearch Bench (test)
56.28Comprehensiveness
14
DeepResearch Bench II
57.58Information Recall
13
DeepResearch Bench 1.0 (test)
46.45Overall Score
12
DeepResearch Bench Open-Ended
52.09Overall Score
11
DeepResearch Bench 100 PhD-level research tasks
54.25Comprehensiveness
9
DeepResearch Bench (DRB)
50.6Score
8
DeepResearch Bench RACE framework 1.0 (test)
49.71Overall Score
7
DeepResearch Bench stratified 10-query subset 1.0
38.5PRISM
6
DeepResearch Bench online interactive settings
36.44Intent Precision
6
DeepResearch Bench 10-query
38.5PRISM (%)
5
DeepResearch Bench 20 reports (sampled)
95Readability (Win/Tie Rate)
5
DeepResearch Bench offline (test)
2.43Quality Score
4
DeepResearch Bench 100 follow-up instances
15.35Instance Score
3