TEST #1547 RESULTS
09/23/2025 - 8:35 PM
ACCELERATOR
MODEL
62.7
tokens/s
generation
539
ms
time to first token
2387
tokens/s
prompt
652
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD EPYC 7532 32-Core Processor (znver2)
RAM
503.5GB
OS
Linux
Kernel Release
6.8.0-59-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #61~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Tue Apr 15 17:03:15 UTC 2
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
3003
tokens/s
64.2
tokens/s
357
ms
pp4096+tg256
2296
tokens/s
56.9
tokens/s
1.80
sec
pp2048+tg256
2508
tokens/s
62.0
tokens/s
833
ms
pp2048+tg768
2717
tokens/s
61.4
tokens/s
770
ms
pp1024+tg1024
3012
tokens/s
63.4
tokens/s
356
ms
pp1280+tg3072
2895
tokens/s
60.2
tokens/s
458
ms
pp384+tg1152
2069
tokens/s
64.7
tokens/s
201
ms
pp64+tg1024
2186
tokens/s
65.9
tokens/s
44
ms
pp16+tg1536
801
tokens/s
65.1
tokens/s
35
ms