Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Long-Context LLM Inference Prefill Performance
Loading...
0.62
Prefill Latency (ms)
Kascade
-87.0508
504.7271
1,096.505
1,688.2829
Dec 18, 2025
Prefill Latency (ms)
Ratio vs TL
Attention Speedup (FA3)
Attention Speedup (TL)
Updated 1mo ago
Evaluation Results
Method
Method
Links
Prefill Latency (ms)
Ratio vs TL
Attention Speedup (FA3)
Attention Speedup (TL)
Kascade
Seqlen=8192, Topk%=10%...
2025.12
0.62
-
1.23
1.62
FA3
Seqlen=8192, Topk%=10%...
2025.12
0.76
-
-
-
Tilelang (TL)
Seqlen=8192, Topk%=10%...
2025.12
1
-
-
-
Kascade
Seqlen=262144, Topk%=1...
2025.12
408.3
-
2.12
2.57
Kascade
Seqlen=262144, Topk%=3...
2025.12
727.55
-
1.19
1.44
Anchor
Seqlen=262144, Topk%=3...
2025.12
2,192.39
2.09
-
-
Feedback
Search any
task
Search any
task
