Long-context retrieval
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
100Accuracy
9
May 19, 2026
100Accuracy
9
May 19, 2026
100S1 Score
9
Apr 20, 2026
99.5Accuracy (4K Context)
8
Jun 19, 2026
100Accuracy (1024 tokens)
8
May 21, 2026
100Accuracy (1024)
8
May 21, 2026
100Single-1 Acc (1K)
8
Apr 10, 2026
95.73Accuracy
8
Feb 26, 2026
85.4Document Avg
8
Feb 26, 2026
100Score (4k Context)
7
Feb 26, 2026
66.4Accuracy
6
Jul 7, 2026
15.3nDCG@10
6
Jun 23, 2026
99NIAH Score
6
Feb 26, 2026
98.6NIAH Score
6
Feb 26, 2026
100Multi-needle Avg Recall
6
Feb 26, 2026
100Accuracy
5
Mar 12, 2026
99.9Score
5
Feb 26, 2026
60SQuAD Score
3
May 28, 2026
0Delta Accuracy (Δacc)
3
May 21, 2026
100Accuracy (512 tokens)
3
Feb 26, 2026
86.22Score
2
Jun 16, 2026
1Score
2
May 15, 2026
1Score
2
May 15, 2026
1Score
2
May 15, 2026
100NIAH Score
2
Apr 23, 2026