๐Ÿ“š ์„ธํ˜„'s Vault

๐ŸŒ ๋„๋ฉ”์ธ

  • ๐Ÿ”ฎ3D-Vision
  • ๐ŸŽจRendering
  • ๐Ÿค–Robotics
  • ๐Ÿง LLM
  • ๐Ÿ‘๏ธVLM
  • ๐ŸŽฌGenAI
  • ๐ŸฅฝXR
  • ๐ŸŽฎSimulation
  • ๐Ÿ› ๏ธDev-Tools
  • ๐Ÿ’ฐCrypto
  • ๐Ÿ“ˆFinance
  • ๐Ÿ“‹Productivity
  • ๐Ÿ“ฆ๊ธฐํƒ€

๐Ÿ“„ Papers

  • ๐Ÿ“š์ „์ฒด ๋…ผ๋ฌธ172
Home

โฏ

bookmarks

โฏ

llama 32 1b in 4 bit runs at 60 tokssec with mlx swift on my iphone 15 pro its

llama-32-1b-in-4-bit-runs-at-60-tokssec-with-mlx-swift-on-my-iphone-15-pro-its

2024๋…„ 9์›” 26์ผ1 min read

  • LLM
  • inference

Awni Hannun (@awnihannun)

2024-09-26 | โค๏ธ 2732 | ๐Ÿ” 240


Llama 3.2 1B in 4-bit runs at ~60 toks/sec with MLX Swift on my iPhone 15 pro.

Itโ€™s quite good and easily runs on-device: https://x.com/awnihannun/status/1839330067039887622/video/1

๋ฏธ๋””์–ด

video


Tags

domain-llm


๊ทธ๋ž˜ํ”„ ๋ทฐ

  • Awni Hannun (@awnihannun)
  • ๋ฏธ๋””์–ด
  • Tags

๋ฐฑ๋งํฌ

  • domain-LLM

Created with Quartz v4.5.2 ยฉ 2026

  • GitHub
  • Sehyeon Park