TEST #2286 RESULTS
01/14/2026 - 11:35 PM
ACCELERATOR
MODEL
89.7
tokens/s
generation
432
ms
time to first token
2880
tokens/s
prompt
843
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
Intel Xeon W-2133 CPU @ 3.60GHz (skylake-avx512)
RAM
125.4GB
OS
Linux
Kernel Release
6.14.0-37-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #37~24.04.1-Ubuntu SMP PREEMPT_DYNAMIC Thu Nov 20 10:25:38 UTC 2
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
3162
tokens/s
90.7
tokens/s
334
ms
pp4096+tg256
2881
tokens/s
79.1
tokens/s
1.43
sec
pp2048+tg256
3321
tokens/s
85.4
tokens/s
629
ms
pp2048+tg768
3373
tokens/s
88.1
tokens/s
619
ms
pp1024+tg1024
3637
tokens/s
92.3
tokens/s
292
ms
pp1280+tg3072
3551
tokens/s
87.0
tokens/s
371
ms
pp384+tg1152
3519
tokens/s
93.6
tokens/s
119
ms
pp64+tg1024
2026
tokens/s
96.0
tokens/s
41
ms
pp16+tg1536
446
tokens/s
95.1
tokens/s
46
ms