Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Long-context evaluation (Speed & Accuracy) on RULER 128K
Loading...
65.03
RULER Score
FullAttention
58.3324
60.0712
61.81
63.5488
Sep 29, 2025
RULER Score
Time To First Token (TTFT) per second
Time To First Token (TTFT) Speedup
Updated 17d ago
Evaluation Results
Method
Method
Links
RULER Score
Time To First Token (TTFT) per second
Time To First Token (TTFT) Speedup
FullAttention
Backbone=Llama3.1-70B-...
2025.09
65.03
91.84
1
ProxyAttn
Backbone=Llama3.1-70B-...
2025.09
62.23
44.23
2.08
MInference
Backbone=Llama3.1-70B-...
2025.09
60.33
82.9
1.11
XAttention
Backbone=Llama3.1-70B-...
2025.09
58.59
60.57
1.52
Feedback
Search any
task
Search any
task