TEST #2904 RESULTS
02/26/2026 - 11:22 AM
ACCELERATOR
MODEL
36.6
tokens/s
generation
670
ms
time to first token
2060
tokens/s
prompt
483
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
RAM
119.7GB
OS
Linux
Kernel Release
6.17.0-1008-nvidia
Architecture
aarch64
Version
Cosmopolitan 3.9.7 MODE=aarch64; #8-Ubuntu SMP PREEMPT_DYNAMIC Wed Jan 21 17:56:56 UTC 2026
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
2531
tokens/s
37.5
tokens/s
434
ms
pp4096+tg256
1723
tokens/s
33.3
tokens/s
2.41
sec
pp2048+tg256
2192
tokens/s
36.2
tokens/s
961
ms
pp2048+tg768
2196
tokens/s
35.9
tokens/s
962
ms
pp1024+tg1024
2516
tokens/s
37.1
tokens/s
433
ms
pp1280+tg3072
2428
tokens/s
35.1
tokens/s
554
ms
pp384+tg1152
2968
tokens/s
37.8
tokens/s
155
ms
pp64+tg1024
1504
tokens/s
38.4
tokens/s
69
ms
pp16+tg1536
482
tokens/s
38.2
tokens/s
59
ms