TEST #3896 RESULTS
07/12/2026 - 2:02 PM
ACCELERATOR
MODEL
7.3
tokens/s
generation
22.19
sec
time to first token
64
tokens/s
prompt
28
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD Ryzen AI 9 HX 370 w/ Radeon 890M
RAM
30GB
OS
Linux
Kernel Release
7.0.0-22-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #22-Ubuntu SMP PREEMPT_DYNAMIC Mon May 25 15:54:34 UTC 2026
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
66
tokens/s
7.2
tokens/s
15.61
sec
pp4096+tg256
57
tokens/s
6.6
tokens/s
71.92
sec
pp2048+tg256
61
tokens/s
7.1
tokens/s
33.73
sec
pp2048+tg768
60
tokens/s
7.0
tokens/s
34.42
sec
pp1024+tg1024
61
tokens/s
7.3
tokens/s
16.98
sec
pp1280+tg3072
65
tokens/s
7.0
tokens/s
19.79
sec
pp384+tg1152
68
tokens/s
7.6
tokens/s
5.78
sec
pp64+tg1024
69
tokens/s
7.8
tokens/s
1.07
sec
pp16+tg1536
68
tokens/s
7.7
tokens/s
369
ms