Adobe Ships Photoshop 27.7 With On-Device AI

Adobe released Photoshop version 27.7 with an on-device generative AI model for the Remove tool and tighter Firefly integration, according to Adobe release notes and community posts. Adobe's community beta post lists the on-device model download size as approximately 5 GB and sets minimum hardware: Apple Silicon M1 Max or higher, 24 GB RAM, and macOS Tahoe on Mac; Windows requirements include NVIDIA GeForce RTX 30-series or higher with 14 GB VRAM, per Adobe's documentation. The Device option is automatically enabled only on compatible systems and can be switched back to cloud processing, per Adobe's announcement. Editorial analysis: this brings local, offline editing to Photoshop users with capable hardware, but the steep requirements limit practical reach for many professionals and hobbyists.
What happened
Adobe released Photoshop version 27.7, and the update adds an on-device generative AI model option for the Remove tool, per Adobe release notes and the Photoshop community announcement for v27.7. Adobe's community beta post for the on-device model lists a download size of approximately 5 GB and specific minimum system requirements: on Mac, an Apple Silicon M1 Max or higher, 24 GB RAM minimum, and macOS Tahoe (macOS 26) or later; on Windows, an NVIDIA GeForce RTX 30-series GPU or higher with at least 14 GB VRAM, or a discrete GPU with 12 GB VRAM (Intel systems with integrated GPU require 32 GB system RAM). Adobe's community posts and release notes state users can switch between cloud and device processing and that compatibility is checked automatically before download.
Technical details
The community documentation marks the on-device Remove model as GPU intensive and requires Photoshop settings to enable GPU acceleration and the Remove Tool's faster processing flags, per Adobe's beta post. The desktop release notes and community announcement also describe deeper integration between Photoshop and Firefly Boards, letting users share documents directly to Boards from the Share panel and the Export menu and surfacing Firefly Boards as a destination in the Variations panel for moodboarding and iteration.
Editorial analysis - technical context
Local inference for image-editing models is increasingly feasible on high-end consumer hardware, and the 27.7 release follows an industry pattern where vendors offer both cloud and device execution to balance latency, privacy, and compute cost. Companies deploying comparable on-device models typically require large local model files and GPU resources, which raises delivery challenges such as download size management, model updates, and hardware compatibility checks.
Context and significance
Industry observers note that on-device processing reduces dependence on cloud connectivity and can improve privacy by keeping image pixels local, but public reporting on this release emphasizes the practical tradeoff: high memory and GPU requirements exclude many older laptops and mainstream desktops. For practitioners, this change matters chiefly in workflows that demand offline editing or strict data residency; for broader teams and cloud-first users, the cloud option remains available and unchanged, per Adobe's documentation.
What to watch
Monitor Adobe's compatibility checks and rollout notes for any relaxation of minimum requirements or optimized, smaller model variants. Also watch whether Adobe expands on-device model support beyond the Remove tool to other generative or neural filters, and whether model download and update flows are streamlined for managed enterprise deployments. Observers will also look for performance benchmarks and user feedback in Adobe community forums about quality parity between cloud and device processing.
Scoring Rationale
The on-device AI option is a notable product-level advance because it enables offline, local editing and privacy improvements for image workflows. However, the high minimum hardware requirements limit immediate impact for the broader Photoshop user base, keeping this release in the "solid" product-update tier for practitioners.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
