TEST #3193 RESULTS
03/25/2026 - 11:04 AM
ACCELERATOR
MODEL
30.8
tokens/s
generation
763
ms
time to first token
1831
tokens/s
prompt
419
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
RAM
121.6GB
OS
Linux
Kernel Release
6.17.0-1014-nvidia
Architecture
aarch64
Version
Cosmopolitan 3.9.7 MODE=aarch64; #14-Ubuntu SMP PREEMPT_DYNAMIC Tue Mar 17 19:01:40 UTC 2026
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
2223
tokens/s
31.2
tokens/s
493
ms
pp4096+tg256
1513
tokens/s
27.9
tokens/s
2.74
sec
pp2048+tg256
1920
tokens/s
30.4
tokens/s
1.10
sec
pp2048+tg768
1935
tokens/s
30.2
tokens/s
1.09
sec
pp1024+tg1024
2236
tokens/s
31.1
tokens/s
493
ms
pp1280+tg3072
2149
tokens/s
29.5
tokens/s
628
ms
pp384+tg1152
2698
tokens/s
31.8
tokens/s
175
ms
pp64+tg1024
1385
tokens/s
32.5
tokens/s
78
ms
pp16+tg1536
420
tokens/s
32.1
tokens/s
69
ms