Flutter 3.44 was released on 18 May 2026, alongside Dart 3.12. The release notes are long, but two changes are architectural rather than incremental: the Skia renderer is being removed on Android 10 and above, and the Material and Cupertino widget libraries are leaving the core SDK. Either one on its own would define a release. Together they tell you where Flutter is going.
Impeller is now the renderer on modern Android
With the Skia backend removed for Android 10 and above, pure Impeller Vulkan rendering is the default on modern devices. The practical consequence is the end of shader compilation jank — the stutter that appeared the first time an animation ran, because the shader had to be compiled at that moment. Startup-to-interaction gets faster and animation performance becomes more predictable.
Shader jank was Flutter's most visible weakness and the single most common complaint in performance reviews of Flutter apps. If you previously shipped shader warm-up files to work around it, you can remove them. Do verify on older hardware in your device matrix: "Vulkan on Android 10+" covers the vast majority of active devices, but a Vulkan driver on a budget 2020 handset is not the same animal as one on a flagship.
Material and Cupertino are leaving the SDK
The Flutter team is decoupling the two design systems from the core framework, freezing them inside the SDK and moving the widgets into independent packages — material_ui and cupertino_ui. This is the biggest under-the-hood change in 3.44.
The reasoning is sound. Material and Cupertino evolve on Google's and Apple's schedules, not Flutter's, and binding them to the framework release cycle slowed both. Decoupled, they can ship when the design systems change. The cost is a migration for every existing app, and a new question in every architecture review: which design system package are you on, and who is upgrading it? Treat this like any other dependency you now own explicitly rather than inherit implicitly.
Cupertino squircles are real squircles
Flutter introduced RSuperellipse, a shape matching Apple's native rounded corners, which required deep changes in the rendering engine to support continuous curvature. Apple's corners are not circular arcs; they are superellipses, and the difference is the reason Flutter iOS apps have always looked subtly, unplaceably wrong to designers with a good eye. This is a small feature that closes a long-standing credibility gap when you tell a client their Flutter app will feel native on iOS.
Agentic hot reload and the MCP server
The headline developer-experience feature is agentic hot reload: the Dart and Flutter MCP server can locate your running app and trigger a hot reload the moment an AI agent applies a code change. That closes the loop that made AI-assisted Flutter development tedious — the agent edits, you alt-tab, you save, you look. Now the agent sees the result of its own change.
Pair this with the GenUI SDK, built on the open A2UI protocol, which lets an agent compose real Flutter widgets at runtime and generate interactive screens based on what the user is trying to do. That is a genuinely new interaction model rather than a chatbot bolted onto a sidebar. It is also early: I would build a prototype before I would build a product on it.
Dart 3.12 and where Flutter is running
Dart 3.12 adds private named parameters and experimental support for primary constructors — the kind of quiet ergonomics that make model classes shorter. On the platform side, Flutter now powers the infotainment system in the 2026 Toyota RAV4, and an official webOS SDK is coming for LG smart TVs. Flutter has been "multi-platform" in marketing for years; shipping in a production car dashboard is a different class of claim.
What to do with an existing Flutter app
Upgrade in stages. Take 3.44 and confirm rendering on your real device matrix, paying attention to custom shaders and platform views, which are where Impeller differences historically surfaced. Delete any shader warm-up workaround and measure whether jank actually went away. Then plan the material_ui migration as its own piece of work with its own testing pass — a design-system swap touches every screen, and no amount of care makes that a low-risk change to bundle with a feature release. Leave GenUI for a spike.
Is Flutter still the right choice in 2026?
For most cross-platform projects, yes, and 3.44 strengthens the case. Impeller removes the performance objection that dominated the argument. The Material decoupling is short-term migration pain in exchange for design systems that no longer lag their native counterparts. The honest counterpoint is unchanged: if your app is a thin client over platform-specific capability — deep HealthKit or ARKit work, tight Wear OS integration — native still wins, and Flutter's expanding platform list does not change that calculus.