TEST #3100 RESULTS
03/16/2026 - 6:23 PM
ACCELERATOR
MODEL
21.4
tokens/s
generation
5.66
sec
time to first token
239
tokens/s
prompt
97
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S
RAM
31.1GB
OS
Linux
Kernel Release
6.17.0-19-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #19~24.04.2-Ubuntu SMP PREEMPT_DYNAMIC Fri Mar 6 23:08:46 UTC 2
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
254
tokens/s
22.6
tokens/s
4.08
sec
pp4096+tg256
218
tokens/s
19.0
tokens/s
18.82
sec
pp2048+tg256
246
tokens/s
21.0
tokens/s
8.39
sec
pp2048+tg768
242
tokens/s
20.6
tokens/s
8.52
sec
pp1024+tg1024
257
tokens/s
21.6
tokens/s
4.03
sec
pp1280+tg3072
254
tokens/s
20.2
tokens/s
5.08
sec
pp384+tg1152
249
tokens/s
22.3
tokens/s
1.59
sec
pp64+tg1024
250
tokens/s
22.8
tokens/s
299
ms
pp16+tg1536
185
tokens/s
22.5
tokens/s
129
ms