TEST #56 RESULTS
03/28/2025 - 8:45 PM
ACCELERATOR
MODEL
77.6
tokens/s
generation
522
ms
time to first token
2484
tokens/s
prompt
718
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD EPYC 7352 24-Core Processor (znver2)
RAM
251.6GB
OS
Linux
Kernel Release
6.8.0-45-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #45~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Wed Sep 11 15:25:05 UTC 2
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
2916
tokens/s
81.5
tokens/s
364
ms
pp4096+tg256
2364
tokens/s
70.1
tokens/s
1.75
sec
pp2048+tg256
2717
tokens/s
77.1
tokens/s
767
ms
pp2048+tg768
2718
tokens/s
76.0
tokens/s
767
ms
pp1024+tg1024
2911
tokens/s
79.0
tokens/s
364
ms
pp1280+tg3072
2809
tokens/s
73.7
tokens/s
469
ms
pp384+tg1152
2980
tokens/s
79.7
tokens/s
142
ms
pp64+tg1024
2368
tokens/s
80.9
tokens/s
39
ms
pp16+tg1536
576
tokens/s
80.4
tokens/s
39
ms