TEST #139 RESULTS
03/28/2025 - 10:18 PM
ACCELERATOR
MODEL
92.6
tokens/s
generation
409
ms
time to first token
3140
tokens/s
prompt
892
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD Ryzen 9 5900X 12-Core Processor (znver3)
RAM
31.3GB
OS
Linux
Kernel Release
6.5.0-44-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #44~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Tue Jun 18 14:36:16 UTC 2
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
3761
tokens/s
98.4
tokens/s
283
ms
pp4096+tg256
3019
tokens/s
85.4
tokens/s
1.37
sec
pp2048+tg256
3481
tokens/s
93.5
tokens/s
599
ms
pp2048+tg768
3473
tokens/s
92.1
tokens/s
600
ms
pp1024+tg1024
3750
tokens/s
94.9
tokens/s
283
ms
pp1280+tg3072
3532
tokens/s
85.7
tokens/s
373
ms
pp384+tg1152
3774
tokens/s
93.1
tokens/s
112
ms
pp64+tg1024
2762
tokens/s
95.4
tokens/s
33
ms
pp16+tg1536
707
tokens/s
94.3
tokens/s
32
ms