M5 Pro
A high-performance chip released by Apple as part of the M5 series. It integrates a neural accelerator and is designed for professional workloads.
Timeline (last 90 days)
This batch of touchscreen OLED MacBook Pro models uses the existing M5 Pro and M5 chips and brings no new processor upgrades.
The Mac version of Locally Uncensored is currently in beta and has only been tested on one M5 Pro (48GB memory).
On an M5 Pro (48 GB memory), LTX-2.3 Distilled 4-bit takes 64 seconds to generate a 2-second 768x512 video clip with sound.
Tested on an M5 Pro (48 GB memory), DreamShaper XL Turbo takes 33 seconds to generate a single image, and Qwen-Image 2.1 Turbo takes 38 seconds.
In the reproduction experiment (torch 2.14.0, macOS 26.5.2, M5 Pro 48 GB, 4 GiB tensor), right after moving to CPU both phys_footprint and the MPS driver count returned to near zero, and only after the CPU read through the copy page by page did the footprint return to 4.2 GiB, with the results of three runs completely identical.
In a run of Qwen-Image-2.1 with diffusers model CPU offload, the footprint on a 48 GB M5 Pro never exceeded 18.5 GB, while swap space grew by 9.6 GB.
Geekerwan determined that the standard M6 is far inferior to the M5 Pro in AI inference due to a large gap in memory bandwidth.
The M5 Pro's GPU bandwidth and GPU scale are somewhat inferior to those of the M5 Max.
Geekerwan recommends the M5 Pro as the most suitable choice when deploying small-to-medium-scale models such as DeepSeek-V4-Flash on Apple's M5 series.
最新款Mac mini搭载M6与M5 Pro芯片，机身为5英寸见方
2 more timeline events
All Facts (8)
This batch of touchscreen OLED MacBook Pro models uses the existing M5 Pro and M5 chips and brings no new processor upgrades.
50%UnverifiedThe M5 Pro's GPU bandwidth and GPU scale are somewhat inferior to those of the M5 Max.
50%UnverifiedGeekerwan recommends the M5 Pro as the most suitable choice when deploying small-to-medium-scale models such as DeepSeek-V4-Flash on Apple's M5 series.
50%UnverifiedGeekerwan determined that the standard M6 is far inferior to the M5 Pro in AI inference due to a large gap in memory bandwidth.
50%Unverified新款Mac mini提供M6标准版和M5 Pro两种芯片选项，M6面向日常用户，M5 Pro面向多任务、大型代码编译及本地AI模型等专业场景
50%UnverifiedIn the reproduction experiment (torch 2.14.0, macOS 26.5.2, M5 Pro 48 GB, 4 GiB tensor), right after moving to CPU both phys_footprint and the MPS driver count returned to near zero, and only after the CPU read through the copy page by page did the footprint return to 4.2 GiB, with the results of three runs completely identical.
50%Unverified最新款Mac mini搭载M6与M5 Pro芯片，机身为5英寸见方
50%UnverifiedM5 Pro预计支持高达48GB或更多统一内存，理论上可在本地运行70亿参数甚至更大规模的量化模型
50%