TEST #55 RESULTS
03/28/2025 - 8:45 PM
ACCELERATOR
MODEL
89.9
tokens/s
generation
445
ms
time to first token
2890
tokens/s
prompt
836
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD EPYC 7B13 64-Core Processor (znver3)
RAM
251.5GB
OS
Linux
Kernel Release
6.8.0-48-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #48~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Mon Oct 7 11:24:13 UTC 2
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
3405
tokens/s
94.3
tokens/s
311
ms
pp4096+tg256
2788
tokens/s
80.3
tokens/s
1.48
sec
pp2048+tg256
3173
tokens/s
88.7
tokens/s
657
ms
pp2048+tg768
3169
tokens/s
87.7
tokens/s
657
ms
pp1024+tg1024
3396
tokens/s
91.7
tokens/s
312
ms
pp1280+tg3072
3302
tokens/s
84.9
tokens/s
398
ms
pp384+tg1152
3390
tokens/s
92.7
tokens/s
124
ms
pp64+tg1024
2700
tokens/s
94.8
tokens/s
33
ms
pp16+tg1536
687
tokens/s
93.8
tokens/s
33
ms