TEST #3972 RESULTS
07/30/2026 - 7:32 PM
ACCELERATOR
MODEL
58.4
tokens/s
generation
1.72
sec
time to first token
706
tokens/s
prompt
288
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
Apple M4 Max 12P+4E
RAM
128GB
OS
Darwin
Kernel Release
25.5.0
Architecture
arm64
Version
Cosmopolitan 3.9.7 MODE=aarch64; Darwin Kernel Version 25.5.0: Tue Jun 9 22:28:34 PDT 2026; root:xnu-12377.121.10~1/RELEASE_ARM64_T6041
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
818
tokens/s
63.7
tokens/s
1.27
sec
pp4096+tg256
755
tokens/s
44.5
tokens/s
5.45
sec
pp2048+tg256
798
tokens/s
55.4
tokens/s
2.58
sec
pp2048+tg768
798
tokens/s
53.7
tokens/s
2.59
sec
pp1024+tg1024
813
tokens/s
59.8
tokens/s
1.27
sec
pp1280+tg3072
807
tokens/s
51.4
tokens/s
1.60
sec
pp384+tg1152
798
tokens/s
64.3
tokens/s
496
ms
pp64+tg1024
557
tokens/s
67.6
tokens/s
129
ms
pp16+tg1536
205
tokens/s
65.3
tokens/s
92
ms