Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

NeedleBench

Benchmarks

Task NameDataset NameSOTA ResultTrend
Multi-objective retrievalNeedleBench M-RT (32k context stress slice)
Slot-set Exactness100
20
Long-context retrieval and reasoningNeedleBench
Accuracy @ 8k Context91.9
5
Showing 2 of 2 rows