Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Latency Evaluation on NSA Workload Head Dimension 128
Loading...
0.67
Latency (s)
LLM-TL
-0.3548
6.5626
13.48
20.3974
Jun 14, 2025
Latency (s)
Speedup
Updated 4d ago
Evaluation Results
Method
Method
Links
Latency (s)
Speedup
LLM-TL
Sequence Length=512, H...
2025.06
0.67
1.25
Naive NSA
Sequence Length=512, H...
2025.06
0.84
-
LLM-TL
Sequence Length=1k, He...
2025.06
1.26
1.33
Naive NSA
Sequence Length=1k, He...
2025.06
1.68
-
LLM-TL
Sequence Length=2k, He...
2025.06
2.59
1.29
Naive NSA
Sequence Length=2k, He...
2025.06
3.35
-
LLM-TL
Sequence Length=4k, He...
2025.06
5.25
1.26
Naive NSA
Sequence Length=4k, He...
2025.06
6.61
-
LLM-TL
Sequence Length=8k, He...
2025.06
10.59
1.26
Naive NSA
Sequence Length=8k, He...
2025.06
13.34
-
LLM-TL
Sequence Length=16k, H...
2025.06
21.27
1.24
Naive NSA
Sequence Length=16k, H...
2025.06
26.29
-
Feedback
Search any
task
Search any
task