Skip to content

[Blog] Benchmarking Prefill–Decode: fixed 1:3 as a strong default #296

[Blog] Benchmarking Prefill–Decode: fixed 1:3 as a strong default

[Blog] Benchmarking Prefill–Decode: fixed 1:3 as a strong default #296

Job Run time
7s
1m 52s
11s
4m 0s
53s
15s
17s
3m 57s
3m 7s
3m 3s
4m 15s
2m 42s
4m 10s
2m 37s
2m 6s
2m 10s
3m 53s
1m 42s
2m 30s
2m 31s
2m 11s
3m 52s
25s
29s
43s
17s
17s
6m 27s
13s
1h 1m 12s