TEST #3650 RESULTS
05/29/2026 - 12:22 PM
ACCELERATOR
MODEL
12.2
tokens/s
generation
5.44
sec
time to first token
251
tokens/s
prompt
83
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD Ryzen 9 9950X3D2 16-Core Processor
RAM
60.5GB
OS
Linux
Kernel Release
7.0.2-6-pve
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #1 SMP PREEMPT_DYNAMIC PMX 7.0.2-6 (2026-05-20T08:55Z)
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
279
tokens/s
12.6
tokens/s
3.76
sec
pp4096+tg256
224
tokens/s
11.3
tokens/s
18.40
sec
pp2048+tg256
254
tokens/s
12.1
tokens/s
8.13
sec
pp2048+tg768
255
tokens/s
12.0
tokens/s
8.13
sec
pp1024+tg1024
273
tokens/s
12.3
tokens/s
3.84
sec
pp1280+tg3072
274
tokens/s
11.8
tokens/s
4.75
sec
pp384+tg1152
284
tokens/s
12.5
tokens/s
1.43
sec
pp64+tg1024
256
tokens/s
12.8
tokens/s
327
ms
pp16+tg1536
157
tokens/s
12.7
tokens/s
179
ms