forked from tabletuser-blogspot/ollama-benchmark
-
Notifications
You must be signed in to change notification settings - Fork 14
Open
Description
Tests from a Macbook Pro M2 Max 16", but there's a Mac studio ~equivalent of this.
- CPU: Total Number of Cores: 12 (8 performance and 4 efficiency)
- GPU: Total Number of Cores: 38
- Memory: 32 GB
- Idling around 18-25 W
Power meter: Athom smart plug
-/+:. root@MATE-MBP
:++++. OS: 64bit Mac OS X 15.1.1 24B91
/+++/. Kernel: arm64 Darwin 24.1.0
.:-::- .+/:-``.::- Uptime: 4d 21h 9m
.:/++++++/::::/++++++/:` Packages: 50
.:///////////////////////:` Shell: bash 5.2.32
////////////////////////` Resolution: 3456x2234
-+++++++++++++++++++++++` DE: Aqua
/++++++++++++++++++++++/ WM: Quartz Compositor
/sssssssssssssssssssssss. WM Theme: Blue
:ssssssssssssssssssssssss- Disk: 11G / 995G (2%)
osssssssssssssssssssssssso/` CPU: Apple M2 Max
`syyyyyyyyyyyyyyyyyyyyyyyy+` GPU: Apple M2 Max
`ossssssssssssssssssssss/ RAM: 4022MiB / 32768MiB
:ooooooooooooooooooo+.
`:+oo+/:-..-:/+o+/-
Here's the markdown table representation:
| Model | Speed (tokens/s) | Power Consumption (W) |
|---|---|---|
| llama3.2:3b | 90.36 | 82-84 |
| llama3.1:8b | 55.67 | 84-88 |
| llama2:13b | 37.17 | 89-91 |
Proofs
Llama3.2:3b
llama3.1:8b
llama2:13b

Metadata
Metadata
Assignees
Labels
No labels