Apple M3 Max vs M5 Max for AI — I Was Wrong

I bought the M5 Max MacBook Pro to run local LLMs faster. Then I ran it
side-by-side against my M3 Max — and I wasn't sure I made the right call.
In this video I bench both chips running llama.cpp with the
same models, same prompts, same context window. No cloud. No API keys. Just your machine doing the heavy lifting. #opensource #llm #m5 #llamacpp #qwen3 #m3 #gemma4
👉ⓢⓤⓑⓢⓒⓡⓘⓑⓔ
👉 !! try Wan Video online at https://agireact.com/wan-t2v !!

If you found this useful:
👍 Like if the results surprised you
🔔 Subscribe for more local AI benchmarks and hardware deep-dives
💬 Drop your setup in the comments — curious what you're running models on

🖥️ Hardware Tested:
- Apple MacBook (unified memory 36GB M3 Max)
- Apple MacBook (unified memory 128GB M5 Max)

🤖 Models Benchmarked:
- Qwen3.6-35B-A3B (Q4)
- Gemma4 31B (Q4)

📄 Qwen3.6 Blog: https://qwen.ai/blog?id=qwen3.6-27b
For Gemma4 model comparison, see https://youtu.be/VYc47oqBnqI

Please join the discord server at https://discord.gg/SgmBydQ2Mn
If you would like to support me, here is my Kofi link: https://ko-fi.com/techpractice and Patreon page: https://www.patreon.com/user?u=89548519
Thank you for watching!

Tutorial links:
For python virtualenv install, see https://youtu.be/uOCL6h9fuVc
ComfyUI for more advanced workflows
ComfyUI on Macbook tutorial: https://youtu.be/ZCswfm0dBYY
FLUX on Macbook: https://youtu.be/asngm4s_9Ho
The ComfyUI workflow can be downloaded from https://github.com/ttio2tech/ComfyUI_workflows_collection (Pulid_flux_workflow.json)

Affiliate links: buy hardware on Amazon
Mac-Mini M4: https://amzn.to/4emPxrB (also has coupon)
AMD GPU: https://amzn.to/3vCp6h1
4600G https://amzn.to/45LhGFa
5600G: https://amzn.to/3LgnFtC (same iGPU, better CPU)
5700G: https://amzn.to/3Z9gUiM (better iGPU, and better CPU)
ssd drive: https://amzn.to/3MVJdg2
DDR4 drive: https://amzn.to/3sKNufi
AM4 motherboard: https://amzn.to/3GfrPit
PSU (power supply unit): https://amzn.to/3Gd87UA
PC Case: https://amzn.to/3QPDNnF
if you are interested in discrete GPU: https://amzn.to/3QT1wDp

Видео Apple M3 Max vs M5 Max for AI — I Was Wrong канала Tech-Practice