TEST #3296 RESULTS
04/12/2026 - 3:00 AM
ACCELERATOR
MODEL
25.1
tokens/s
generation
508
ms
time to first token
2540
tokens/s
prompt
501
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD Ryzen 7 7800X3D 8-Core Processor (znver4)
RAM
31.1GB
OS
Windows
Kernel Release
10.0
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64
RUNTIME
Name
llamafile
Version
0.9.3
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
3258
tokens/s
26.5
tokens/s
352
ms
pp4096+tg256
2601
tokens/s
24.3
tokens/s
1.62
sec
pp2048+tg256
3038
tokens/s
25.5
tokens/s
714
ms
pp2048+tg768
3047
tokens/s
25.2
tokens/s
712
ms
pp1024+tg1024
3324
tokens/s
25.5
tokens/s
348
ms
pp1280+tg3072
3081
tokens/s
24.8
tokens/s
455
ms
pp384+tg1152
3111
tokens/s
25.2
tokens/s
170
ms
pp64+tg1024
1102
tokens/s
24.1
tokens/s
97
ms
pp16+tg1536
299
tokens/s
24.7
tokens/s
101
ms