TEST #3523 RESULTS
05/13/2026 - 6:07 PM
ACCELERATOR
MODEL
64.9
tokens/s
generation
656
ms
time to first token
1891
tokens/s
prompt
572
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD EPYC 7251 8-Core Processor
RAM
220.3GB
OS
Linux
Kernel Release
6.17.0-23-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #23~24.04.1-Ubuntu SMP PREEMPT_DYNAMIC Tue Apr 14 16:11:48 UTC 2
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
2143
tokens/s
66.6
tokens/s
492
ms
pp4096+tg256
1908
tokens/s
53.8
tokens/s
2.17
sec
pp2048+tg256
2183
tokens/s
62.2
tokens/s
954
ms
pp2048+tg768
2185
tokens/s
61.6
tokens/s
953
ms
pp1024+tg1024
2323
tokens/s
66.5
tokens/s
455
ms
pp1280+tg3072
2266
tokens/s
59.8
tokens/s
580
ms
pp384+tg1152
2281
tokens/s
69.6
tokens/s
182
ms
pp64+tg1024
1371
tokens/s
72.5
tokens/s
60
ms
pp16+tg1536
361
tokens/s
71.1
tokens/s
58
ms