IOS 27 RAW 9 Engine Reprocesses Old Photos With Neural Chip

Apple’s RAW processing engine has handled files from hundreds of third-party cameras for years, quietly sitting inside Core Image where most users never think about it. The upgrade arriving with iOS 27 and macOS 27 makes that invisible layer considerably more capable.

The new engine, called RAW 9, uses a tiled CoreML model that combines demosaic and denoise processing in a single pass. The Apple Neural Engine runs the model entirely on device, which means apps using Core Image get improved RAW rendering without any cloud dependency. Apple describes it as the biggest update yet to its RAW processing system.

The practical improvements break down across a few areas:

• Luma and chroma noise reduction while preserving frame detail
• Better demosaicing for color accuracy and separation
• Improved handling of high-ISO files from Canon, Sony, and Fujifilm cameras

What Apple is not leading with is the retroactive benefit. RAW 9 can reprocess photos already sitting in a library, meaning images shot years ago on a mirrorless camera could render differently after the update. Apple’s own examples show clearer text, sharper textures, and cleaner highlights in difficult, noisy files.

What this means for the hardware pipeline

The on-device processing angle matters more as Apple continues pushing the Neural Engine as a workhorse. The A20 chip expected in iPhone 18 carries estimated performance gains over its predecessor, and a more demanding CoreML model like RAW 9 is exactly the kind of workload that benefits from that headroom. Faster inference means faster RAW rendering in photo editing apps without draining the battery.

For developers, Apple has provided guidance on enabling RAW 9 and improving export performance through Core Image. Apps that already use the system-level pipeline get the upgrade without rebuilding from scratch, which lowers the barrier for third-party editors to adopt it.