TEST #85 RESULTS
03/28/2025 - 9:20 PM
ACCELERATOR
MODEL
45.1
tokens/s
generation
274
ms
time to first token
5065
tokens/s
prompt
941
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD EPYC 9354 32-Core Processor (znver4)
RAM
755.6GB
OS
Linux
Kernel Release
6.8.0-50-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #51~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Thu Nov 21 12:03:03 UTC 2
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
7019
tokens/s
47.6
tokens/s
168
ms
pp4096+tg256
4340
tokens/s
38.7
tokens/s
973
ms
pp2048+tg256
5712
tokens/s
43.6
tokens/s
382
ms
pp2048+tg768
5740
tokens/s
42.5
tokens/s
380
ms
pp1024+tg1024
7067
tokens/s
45.3
tokens/s
167
ms
pp1280+tg3072
6311
tokens/s
41.5
tokens/s
225
ms
pp384+tg1152
6877
tokens/s
47.8
tokens/s
76
ms
pp64+tg1024
1925
tokens/s
50.3
tokens/s
53
ms
pp16+tg1536
590
tokens/s
49.1
tokens/s
46
ms