TEST #3747 RESULTS
06/17/2026 - 2:43 PM
ACCELERATOR
MODEL
46.8
tokens/s
generation
1.43
sec
time to first token
889
tokens/s
prompt
307
LocalScore
HOW YOU STACK UP
Explore All ResultsMeta Llama 3.1 8B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD Ryzen Threadripper PRO 3945WX 12-Cores (znver2)
RAM
123.4GB
OS
Linux
Kernel Release
7.0.0-22-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #22-Ubuntu SMP PREEMPT_DYNAMIC Mon May 25 15:54:34 UTC 2026
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
990
tokens/s
56.5
tokens/s
1.05
sec
pp4096+tg256
849
tokens/s
13.2
tokens/s
4.90
sec
pp2048+tg256
1002
tokens/s
42.3
tokens/s
2.07
sec
pp2048+tg768
1003
tokens/s
40.5
tokens/s
2.07
sec
pp1024+tg1024
1089
tokens/s
49.5
tokens/s
959
ms
pp1280+tg3072
1046
tokens/s
32.1
tokens/s
1.24
sec
pp384+tg1152
1138
tokens/s
58.7
tokens/s
352
ms
pp64+tg1024
378
tokens/s
66.3
tokens/s
183
ms
pp16+tg1536
503
tokens/s
61.8
tokens/s
44
ms