Rare find

Core AI Models. Tools to build AI that runs on your phone or Mac without sending data to the cloud.

github.com/apple/coreai-models

Problem

The full read on this project is a sign-in away — what it solves, who it serves, and where it sits in the market.

Audience

The full read on this project is a sign-in away — what it solves, who it serves, and where it sits in the market.

Market

The full read on this project is a sign-in away — what it solves, who it serves, and where it sits in the market.

Pricing

The full read on this project is a sign-in away — what it solves, who it serves, and where it sits in the market.

Vaya's read on this project

Problem, audience, market, and the verdict — sign in to see it.

Updates

July 2026

Move away from deprecated FM API (#123) Jul 23
Polish a few APIs, method names, and remove unnecessarily vending pu… Jul 23
Fix pipelined sampling corruption: use per-call execution descriptor … Jul 22
Fix guided generation: stop on grammar termination, include <|endofte… Jul 22
Flux2 Updates to Improve Image Quality (#120) Jul 21
Stop pipelined generation when consumer drops the stream (#113) Jul 17
Add configurable VLM image preprocessing strategy (#108) Jul 17
Fix diffusion GPU memory leak: reuse InferenceFunction (#110) Jul 17
Fix Qwen3-VL normalization: use 0.5/0.5/0.5 from checkpoint (#105) Jul 16
Fix SD text encoder crash: infer sequence length from model (#103) Jul 16
Adding SAM3 iOS Model (#106) Jul 15
VLM: fix EmbeddedInput shape handling, error recovery, and parallel l… Jul 15
VLM support for FoundationModels protocol (#97) Jul 14
Enable running `iOS` and `macOS` authored models on CUDA GPUs (#99) Jul 14
Fresh export needs tokenizer patterns (#95) Jul 10
Add AIModel metadata for Qwen3-VL-2B-Instruct (#94) Jul 10
change object detector input array creation to use span (#93) Jul 9
Lazy runner design: defer engine load (#91) Jul 9
Release —v0.2.0 Jul 8
fix llm runner main (#90) Jul 8
Use views in consuming manner, not mutating (#89) Jul 8