TEST #682 RESULTS
04/10/2025 - 7:37 PM
ACCELERATOR
MODEL
121
tokens/s
generation
199
ms
time to first token
6808
tokens/s
prompt
1605
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD Ryzen Threadripper PRO 5975WX 32-Cores (znver3)
RAM
503.5GB
OS
Linux
Kernel Release
6.8.0-52-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #53-Ubuntu SMP PREEMPT_DYNAMIC Sat Jan 11 00:06:25 UTC 2025
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
8306
tokens/s
120
tokens/s
132
ms
pp4096+tg256
6096
tokens/s
113
tokens/s
681
ms
pp2048+tg256
7142
tokens/s
118
tokens/s
296
ms
pp2048+tg768
7076
tokens/s
118
tokens/s
299
ms
pp1024+tg1024
9441
tokens/s
122
tokens/s
117
ms
pp1280+tg3072
7627
tokens/s
117
tokens/s
177
ms
pp384+tg1152
9319
tokens/s
125
tokens/s
49
ms
pp64+tg1024
4771
tokens/s
129
tokens/s
21
ms
pp16+tg1536
1493
tokens/s
127
tokens/s
18
ms