Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
LLM Generation Throughput (End-to-end generation)
Loading...
751
Throughput (tokens/s)
STAR-KV
-11.32
186.59
384.5
582.41
Jun 7, 2026
Throughput (tokens/s)
Relative Gain
Updated 1mo ago
Evaluation Results
Method
Method
Links
Throughput (tokens/s)
Relative Gain
STAR-KV
Context Length=1K, Bat...
2026.06
751
3.01
STAR-KV
Context Length=2K, Bat...
2026.06
400.7
3.07
Pytorch SDPA
Context Length=1K, Bat...
2026.06
249.8
-
STAR-KV
Context Length=4K, Bat...
2026.06
212.2
3.07
Pytorch SDPA
Context Length=2K, Bat...
2026.06
130.4
-
STAR-KV
Context Length=8K, Bat...
2026.06
110.6
3.11
Pytorch SDPA
Context Length=4K, Bat...
2026.06
69
-
STAR-KV
Context Length=16K, Ba...
2026.06
56.6
3.14
Pytorch SDPA
Context Length=8K, Bat...
2026.06
35.6
-
Pytorch SDPA
Context Length=16K, Ba...
2026.06
18
-
Feedback
Search any
task
Search any
task