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

S-NIAH

Benchmarks

Task NameDataset NameSOTA ResultTrend
Long-context retrievalS-NIAH
Latency (s)10.5
27
Semantic Needle-In-A-HaystackS-NIAH
Accuracy52.4
27
Number in haystack retrievalS-NIAH-2 number
Accuracy (1K Context)100
14
Single-Needle-in-a-HaystackS-NIAH (test)
Exact Match Accuracy100
12
Long-context RetrievalS-NIAH
Exact Match Accuracy99.6
12
Synthetic RetrievalS-NIAH
Score98.9
4
Showing 6 of 6 rows