TEST #3620 RESULTS
05/26/2026 - 4:39 PM
ACCELERATOR
MODEL
377
tokens/s
generation
84
ms
time to first token
15607
tokens/s
prompt
4117
LocalScore
HOW YOU STACK UP
Explore All ResultsLlama 3.2 1B Instruct - Q4_K - Medium
SYSTEM
CPU
AMD EPYC 7B13 64-Core Processor (znver3)
RAM
125.6GB
OS
Linux
Kernel Release
6.8.0-117-generic
Architecture
x86_64
Version
Cosmopolitan 3.9.7 MODE=x86_64; #117~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Thu May 7 22:17:46 UTC
RUNTIME
Name
llamafile
Version
0.9.2
Commit Hash
a30b324
DETAILED RESULTS
TEST NAME
PROMPT
GENERATION
TTFT
pp1024+tg16
17233
tokens/s
390
tokens/s
62
ms
pp4096+tg256
14053
tokens/s
324
tokens/s
295
ms
pp2048+tg256
18021
tokens/s
371
tokens/s
116
ms
pp2048+tg768
17447
tokens/s
363
tokens/s
120
ms
pp1024+tg1024
20248
tokens/s
383
tokens/s
53
ms
pp1280+tg3072
19439
tokens/s
352
tokens/s
68
ms
pp384+tg1152
22153
tokens/s
394
tokens/s
20
ms
pp64+tg1024
10627
tokens/s
410
tokens/s
8
ms
pp16+tg1536
1242
tokens/s
404
tokens/s
15
ms