Back to browse
GitHub Repository

Community model zoo for Apple Core AI (iOS/macOS 27): 49 models — LLM, VLM, OCR, ASR, TTS, image/video/music gen, forecasting — converted, verified on real devices, downloadable from Hugging Face, and runnable in one line of Swift via CoreAIKit. Plus conversion recipes, on-device benchmarks, custom Metal kernels, and a knowledge base.

353 starsPython

I ported 11 model families to Apple's new on-device AI framework

by mlboy·Jun 12, 2026·3 points·2 comments

AI Analysis

●●●BangerWizardryBig BrainNiche Gem

11 model families ported to Core AI with verified on-device benchmarks on iPhone 17 Pro.

Strengths
  • Custom Metal kernels and conversion code enable genuine on-device inference.
  • Comprehensive benchmark table with tok/s across GPU and ANE.
  • Vision-language models (Qwen3-VL, Gemma 4 VL) included, not just text.
Weaknesses
  • Requires iOS 27/macOS 27, limiting current adoption.
  • Large models (35B MoE) Mac-only, not portable to iPhone.
Category
Target Audience

iOS/macOS developers, on-device AI practitioners, Apple Silicon users

Similar To

Off Grid · MLC Chat · Llamafile

Similar Projects