TEST #3184 RESULTS
03/25/2026 - 12:46 AM
ACCELERATOR
MODEL
62.2
tokens/s
generation
1.94
sec
time to first token
635
tokens/s
prompt
273
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
Apple M5 Max 6P+12E
RAM
36GB
OS
Darwin
Kernel Release
25.3.0
Architecture
arm64
Version
Cosmopolitan 3.9.7 MODE=aarch64; Darwin Kernel Version 25.3.0: Wed Jan 28 20:54:38 PST 2026; root:xnu-12377.91.3~2/RELEASE_ARM64_T6050
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
700
tokens/s
62.0
tokens/s
1.48
sec
pp4096+tg256
654
tokens/s
52.8
tokens/s
6.29
sec
pp2048+tg256
710
tokens/s
60.7
tokens/s
2.90
sec
pp2048+tg768
713
tokens/s
59.5
tokens/s
2.89
sec
pp1024+tg1024
739
tokens/s
63.7
tokens/s
1.40
sec
pp1280+tg3072
733
tokens/s
58.1
tokens/s
1.76
sec
pp384+tg1152
740
tokens/s
66.8
tokens/s
533
ms
pp64+tg1024
530
tokens/s
68.9
tokens/s
135
ms
pp16+tg1536
198
tokens/s
67.4
tokens/s
95
ms