TEST #3961 RESULTS
07/28/2026 - 6:04 PM
ACCELERATOR
MODEL
32.8
tokens/s
generation
3.44
sec
time to first token
358
tokens/s
prompt
151
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
Apple M4 Pro 10P+4E
RAM
48GB
OS
Darwin
Kernel Release
24.6.0
Architecture
arm64
Version
Cosmopolitan 3.9.7 MODE=aarch64; Darwin Kernel Version 24.6.0: Mon Jan 19 22:01:58 PST 2026; root:xnu-11417.140.69.708.3~1/RELEASE_ARM64_T6041
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
417
tokens/s
35.7
tokens/s
2.48
sec
pp4096+tg256
381
tokens/s
25.2
tokens/s
10.80
sec
pp2048+tg256
391
tokens/s
31.1
tokens/s
5.27
sec
pp2048+tg768
399
tokens/s
30.2
tokens/s
5.17
sec
pp1024+tg1024
397
tokens/s
33.5
tokens/s
2.61
sec
pp1280+tg3072
396
tokens/s
29.0
tokens/s
3.26
sec
pp384+tg1152
413
tokens/s
36.1
tokens/s
955
ms
pp64+tg1024
303
tokens/s
37.8
tokens/s
236
ms
pp16+tg1536
123
tokens/s
36.5
tokens/s
154
ms