TEST #54 RESULTS
03/28/2025 - 8:44 PM
ACCELERATOR
MODEL
80.6
tokens/s
generation
401
ms
time to first token
3270
tokens/s
prompt
870
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
Intel Xeon Gold 6342 CPU @ 2.80GHz (icelake-server)
RAM
503.5GB
OS
Linux
Kernel Release
6.8.0-48-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #48~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Mon Oct 7 11:24:13 UTC 2
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
3936
tokens/s
81.9
tokens/s
272
ms
pp4096+tg256
3017
tokens/s
73.8
tokens/s
1.37
sec
pp2048+tg256
3591
tokens/s
80.0
tokens/s
583
ms
pp2048+tg768
3590
tokens/s
79.2
tokens/s
583
ms
pp1024+tg1024
3933
tokens/s
81.6
tokens/s
273
ms
pp1280+tg3072
3794
tokens/s
77.5
tokens/s
350
ms
pp384+tg1152
4216
tokens/s
82.8
tokens/s
103
ms
pp64+tg1024
2711
tokens/s
84.7
tokens/s
35
ms
pp16+tg1536
643
tokens/s
83.9
tokens/s
36
ms