MacBook Pro M5 Max Local LLM Limit Test | Qwen / Llama / DeepSeek / GPT-OSS on MLX ローカルLLMの限界に挑む

Описание: MacBook Pro M5 Max 128GB running local LLMs with MLX.
Testing Qwen 3.5 122B, Llama 3.3 70B, GPT-OSS 120B, DeepSeek R1 70B.
Pushing to the limit.
Shot on iPhone 17 Pro Max

Welcome to the Future
Era of Local LLMs

Great ideas start from here

-Models tested:
・Qwen 3.5 122B-A10B 4bit (Alibaba)
・Llama 3.3 70B 4bit (Meta)
・GPT-OSS 120B 4bit (OpenAI)
・DeepSeek R1 Distill 70B Q8 (DeepSeek)
・Qwen3-235B-A22B

Framework: MLX by Apple
Hardware: MacBook Pro M5 Max 128GB

MacBook Pro M5 Max 128GB でローカルLLMをMLXで動かす実験。
Qwen 3.5 122B、Llama 3.3 70B、GPT-OSS 120B、DeepSeek R1 70B を検証。
未来はインターネット圏外のオフラインでAIが動く。


Ken's Travel Blog
World Odyssey 23 Seconds Round The World
https://worldodyssey.net/


Ken's Tech Blog
Future Odyssey
  / u23ken  


[VIDEO INFO]
Recoding Date: 2026.3.15
Filming location: U23 SOHO
Filmed with: iPhone 17 Pro Max
Edited with: Final Cut Pro 12, MacBook Pro 14 inch M5 Max, Apple Vision Pro
Exported with: Compressor (4K HLG 10 bit)

Produced by KEN UTSUMI

[LINKS]
Instagram:   / u23ken  
X: https://x.com/u23ken/
Travel Blog: www.worldodyssey.net
Website: www.KenUtsumi.com
Apple Books - https://apple.co/2TCcFwz
Kindle - https://www.amazon.co.jp/stores/KEN-U...

[PR]
Contact me for video planning, shooting, and editing around the world.
世界中の動画の企画・撮影・編集の依頼はこちらまで
  / u23ken