TEST #52 RESULTS
03/28/2025 - 8:43 PM
ACCELERATOR
MODEL
90.5
tokens/s
generation
378
ms
time to first token
3447
tokens/s
prompt
938
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD EPYC 7543 32-Core Processor (znver3)
RAM
503.7GB
OS
Linux
Kernel Release
5.4.0-155-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #172-Ubuntu SMP Fri Jul 7 16:10:02 UTC 2023
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
4105
tokens/s
92.6
tokens/s
260
ms
pp4096+tg256
3253
tokens/s
82.5
tokens/s
1.27
sec
pp2048+tg256
3771
tokens/s
89.8
tokens/s
555
ms
pp2048+tg768
3780
tokens/s
89.0
tokens/s
553
ms
pp1024+tg1024
4082
tokens/s
91.7
tokens/s
262
ms
pp1280+tg3072
3963
tokens/s
86.9
tokens/s
334
ms
pp384+tg1152
4292
tokens/s
92.9
tokens/s
100
ms
pp64+tg1024
2994
tokens/s
95.0
tokens/s
31
ms
pp16+tg1536
785
tokens/s
93.8
tokens/s
31
ms