Also a CoreML implementation which seems more efficient than the MLX one. github.com/mizorewww/la...
GitHub - mizorewww/laya-coreml: Local Laya typed decisions on Apple Core ML and Neural Engine. Validated ports, ~5 ms short decisions on M3 Max, reproducible speed and energy benchmarks.
Local Laya typed decisions on Apple Core ML and Neural Engine. Validated ports, ~5 ms short decisions on M3 Max, reproducible speed and energy benchmarks. - mizorewww/laya-coreml
github.com
September 22, 2026 at 7:48 AM
Also a CoreML implementation which seems more efficient than the MLX one. github.com/mizorewww/la...
Released RVM CoreML for Unity v1.0.0
github.com/keijiro/unit...
github.com/keijiro/unit...
September 21, 2026 at 7:58 AM
Released RVM CoreML for Unity v1.0.0
github.com/keijiro/unit...
github.com/keijiro/unit...
This Friday @fosiadesign.bsky.social and I got an app idea we got really excited about. We spent the weekend playing around with RealityKit, CoreML, and ARKit and got a little tech demo working 😊
#visionOS #realityView #coreML #ARKit #buildInPublic #indieDev
#visionOS #realityView #coreML #ARKit #buildInPublic #indieDev
November 24, 2024 at 10:21 PM
This Friday @fosiadesign.bsky.social and I got an app idea we got really excited about. We spent the weekend playing around with RealityKit, CoreML, and ARKit and got a little tech demo working 😊
#visionOS #realityView #coreML #ARKit #buildInPublic #indieDev
#visionOS #realityView #coreML #ARKit #buildInPublic #indieDev
Ha! Here’s a laya that uses coreml! github.com/mizorewww/la...
September 21, 2026 at 7:04 AM
Ha! Here’s a laya that uses coreml! github.com/mizorewww/la...
An open-source version of Jev, called Laya, ran offline on a Mac M4 at about 45 decisions per second using CoreML.
Laya on Mac m4 CoreML Offline https://github.com/mizorewww/laya-coreml
Laya on Mac m4 CoreML Offline https://github.com/mizorewww/laya-coreml - laya.sh
gist.github.com
September 20, 2026 at 10:48 PM
An open-source version of Jev, called Laya, ran offline on a Mac M4 at about 45 decisions per second using CoreML.
エフェクト選択SNS的UI部だけでもクロスプラットフォーム化と、MLをCoreMLからTFLite等で共通化くらいしないと厳しいかなー
September 21, 2026 at 6:09 AM
エフェクト選択SNS的UI部だけでもクロスプラットフォーム化と、MLをCoreMLからTFLite等で共通化くらいしないと厳しいかなー
here's a fun side project i've been working on: i compiled a joint text<>audio embedding model to a fast coreml pipeline, and built a very fast (~400ms for 50k samples, can scale to millions) UMAP dimensionality reduction GPU impl in mlx. using it to browse music libraries and do sample sim search
January 26, 2026 at 6:26 AM
here's a fun side project i've been working on: i compiled a joint text<>audio embedding model to a fast coreml pipeline, and built a very fast (~400ms for 50k samples, can scale to millions) UMAP dimensionality reduction GPU impl in mlx. using it to browse music libraries and do sample sim search
anyone have some Apple silicon to test this? is Jev! already over-throned?
github.com/mizorewww/la...
github.com/mizorewww/la...
September 20, 2026 at 11:24 PM
anyone have some Apple silicon to test this? is Jev! already over-throned?
github.com/mizorewww/la...
github.com/mizorewww/la...
Apple released blazingly fast CoreML models AND an iOS app to run them on iPhone! ⚡
> S0 matches OpenAI's ViT-B/16 in zero-shot performance but is 4.8x faster and 2.8x smaller
> S2 outperforms SigLIP's ViT-B/16 in zero-shot accuracy, being 2.3x faster, 2.1x smaller, and trained with 3x fewer data
> S0 matches OpenAI's ViT-B/16 in zero-shot performance but is 4.8x faster and 2.8x smaller
> S2 outperforms SigLIP's ViT-B/16 in zero-shot accuracy, being 2.3x faster, 2.1x smaller, and trained with 3x fewer data
November 23, 2024 at 4:22 PM
Apple released blazingly fast CoreML models AND an iOS app to run them on iPhone! ⚡
> S0 matches OpenAI's ViT-B/16 in zero-shot performance but is 4.8x faster and 2.8x smaller
> S2 outperforms SigLIP's ViT-B/16 in zero-shot accuracy, being 2.3x faster, 2.1x smaller, and trained with 3x fewer data
> S0 matches OpenAI's ViT-B/16 in zero-shot performance but is 4.8x faster and 2.8x smaller
> S2 outperforms SigLIP's ViT-B/16 in zero-shot accuracy, being 2.3x faster, 2.1x smaller, and trained with 3x fewer data
Mentioned a few times I've been looking at ANE (Apple Neural Engine, the fp16 chip in all new Macs & iPhones).
Just came across another great resource, published a few weeks ago, reverse engineered a lot of the lower level details that CoreML hides from you:
arxiv.org/pdf/2606.22283
Just came across another great resource, published a few weeks ago, reverse engineered a lot of the lower level details that CoreML hides from you:
arxiv.org/pdf/2606.22283
arxiv.org
July 21, 2026 at 6:54 AM
Mentioned a few times I've been looking at ANE (Apple Neural Engine, the fp16 chip in all new Macs & iPhones).
Just came across another great resource, published a few weeks ago, reverse engineered a lot of the lower level details that CoreML hides from you:
arxiv.org/pdf/2606.22283
Just came across another great resource, published a few weeks ago, reverse engineered a lot of the lower level details that CoreML hides from you:
arxiv.org/pdf/2606.22283
Laya on Mac M4 CoreML Offline
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
[comments] [147 points]
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
[comments] [147 points]
Laya on Mac m4 CoreML Offline https://github.com/mizorewww/laya-coreml
Laya on Mac m4 CoreML Offline https://github.com/mizorewww/laya-coreml - laya.sh
gist.github.com
September 21, 2026 at 5:03 AM
Laya on Mac M4 CoreML Offline
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
[comments] [147 points]
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
[comments] [147 points]
The batch that waits for the slowest type — https://q08.org/p/2026-09-24-mizorewww-laya-coreml-wlxlau
September 24, 2026 at 6:31 PM
The batch that waits for the slowest type — https://q08.org/p/2026-09-24-mizorewww-laya-coreml-wlxlau
Laya on Mac M4 CoreML Offline
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
https://news.ycombinator.com/item?id=49777106
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
https://news.ycombinator.com/item?id=49777106
Laya on Mac m4 CoreML Offline https://github.com/mizorewww/laya-coreml
Laya on Mac m4 CoreML Offline https://github.com/mizorewww/laya-coreml - laya.sh
gist.github.com
September 21, 2026 at 1:15 AM
Laya on Mac M4 CoreML Offline
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
https://news.ycombinator.com/item?id=49777106
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
https://news.ycombinator.com/item?id=49777106
🚨 New app drop!
MagicaLCore for iPad lets you create image classifiers using your own dataset — no coding required.
• Build & test CoreML models on-device
• Export .mlmodel in seconds
• Free to try, with 3-day trial
🔗 apps.apple.com/app/id646644...
#indieDev #CoreML #AI #NoCode
MagicaLCore for iPad lets you create image classifiers using your own dataset — no coding required.
• Build & test CoreML models on-device
• Export .mlmodel in seconds
• Free to try, with 3-day trial
🔗 apps.apple.com/app/id646644...
#indieDev #CoreML #AI #NoCode
MagicaLCore
MagicaLCore offers everything you need to work with machine learning on your iPad.
From importing and organizing models to training and live testing, MagicaLCore lets you build and experiment with m...
apps.apple.com
June 23, 2025 at 3:34 PM
🚨 New app drop!
MagicaLCore for iPad lets you create image classifiers using your own dataset — no coding required.
• Build & test CoreML models on-device
• Export .mlmodel in seconds
• Free to try, with 3-day trial
🔗 apps.apple.com/app/id646644...
#indieDev #CoreML #AI #NoCode
MagicaLCore for iPad lets you create image classifiers using your own dataset — no coding required.
• Build & test CoreML models on-device
• Export .mlmodel in seconds
• Free to try, with 3-day trial
🔗 apps.apple.com/app/id646644...
#indieDev #CoreML #AI #NoCode
It’s mental how much boilerplate the HuggingFace libraries have solved. From fine-tuning to exporting to CoreML 🔥
February 8, 2025 at 11:32 AM
It’s mental how much boilerplate the HuggingFace libraries have solved. From fine-tuning to exporting to CoreML 🔥
So far iOS with the same embedding model is faster, by no surprise, in CoreML (38ms) but LiteRT(42ms) is 4ms slower with less code. ONNX is +100ms slower, at 192ms to load the same model, “all-MiniLM-L6-v2”, into memory to start embedding things.
September 23, 2026 at 5:09 PM
So far iOS with the same embedding model is faster, by no surprise, in CoreML (38ms) but LiteRT(42ms) is 4ms slower with less code. ONNX is +100ms slower, at 192ms to load the same model, “all-MiniLM-L6-v2”, into memory to start embedding things.
Laya (OS Jev) on Mac M4 CoreML Offline (45 decisions per second)
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
https://news.ycombinator.com/item?id=49777106
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
https://news.ycombinator.com/item?id=49777106
Laya on Mac m4 CoreML Offline https://github.com/mizorewww/laya-coreml
Laya on Mac m4 CoreML Offline https://github.com/mizorewww/laya-coreml - laya.sh
gist.github.com
September 20, 2026 at 9:15 PM
Laya (OS Jev) on Mac M4 CoreML Offline (45 decisions per second)
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
https://news.ycombinator.com/item?id=49777106
https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0
https://news.ycombinator.com/item?id=49777106
I just finished Day 28 of #100DaysOfSwiftUI Today was my first introduction to CoreML. Some very cool stuff in there.
Day 28 – 100 Days of SwiftUI
Follow the 100 Days of SwiftUI and learn to build apps for free.
www.hackingwithswift.com
November 12, 2024 at 2:03 AM
I just finished Day 28 of #100DaysOfSwiftUI Today was my first introduction to CoreML. Some very cool stuff in there.
SwampMachine V1.0のCoreML変換モデル
Stable Diffusion CoreML ml-stable-diffusion
Mochi Diffusion v5.0
テスト投稿
Stable Diffusion CoreML ml-stable-diffusion
Mochi Diffusion v5.0
テスト投稿
February 11, 2024 at 12:57 AM
SwampMachine V1.0のCoreML変換モデル
Stable Diffusion CoreML ml-stable-diffusion
Mochi Diffusion v5.0
テスト投稿
Stable Diffusion CoreML ml-stable-diffusion
Mochi Diffusion v5.0
テスト投稿
Laya (OS Jev) on Mac M4 CoreML Offline (45 decisions per second)
Laya on Mac m4 CoreML Offline
Laya on Mac m4 CoreML Offline. GitHub Gist: instantly share code, notes, and snippets.
gist.github.com
September 20, 2026 at 6:17 PM
Laya (OS Jev) on Mac M4 CoreML Offline (45 decisions per second)
Laya (OS Jev) on Mac M4 CoreML Offline (45 decisions per second) | Discussion
Laya on Mac m4 CoreML Offline
Laya on Mac m4 CoreML Offline. GitHub Gist: instantly share code, notes, and snippets.
gist.github.com
September 20, 2026 at 6:00 PM
Laya (OS Jev) on Mac M4 CoreML Offline (45 decisions per second) | Discussion
I’ve run Qwen on Apple Silicon before but only with PyTorch MPS backend. Gotta dig into this and see how much of the ANE findings can be translated back into using this via CoreML in an app-store compatible way
March 3, 2026 at 1:48 PM
I’ve run Qwen on Apple Silicon before but only with PyTorch MPS backend. Gotta dig into this and see how much of the ANE findings can be translated back into using this via CoreML in an app-store compatible way
Laya (OS Jev) on Mac M4 CoreML Offline (45 decisions per second) https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0 (https://news.ycombinator.com/item?id=49777106)
Laya on Mac m4 CoreML Offline
Laya on Mac m4 CoreML Offline. GitHub Gist: instantly share code, notes, and snippets.
gist.github.com
September 20, 2026 at 5:13 PM
Laya (OS Jev) on Mac M4 CoreML Offline (45 decisions per second) https://gist.github.com/fordnox/e592d0f68b543fd044be8e6d040863a0 (https://news.ycombinator.com/item?id=49777106)