Flet 1.0 puts synchronous event handlers on the application's event loop

In Flet 0.28, a synchronous event handler went to a thread pool, so a function that read a file or waited on a network response did not hold up the interface. Version 1.0 runs synchronous handlers directly on the application's event loop ^[11]. Blocking work inside a handler now freezes interface updates until the developer converts it to asynchronous code or moves it onto a thread of their own ^[11]. Porting an existing application means reading every handler for calls that block.

Flet came out of Pglet, an earlier web framework designed around several programming languages, and those inherited decisions became a constraint once Flet went Python-centric and added mobile and desktop targets ^[4]. In the June 2025 alpha announcement, Fitsner described mounting technical debt and architectural complexity, then explained that he had chosen a ground-up rewrite while the community was still small enough to absorb breaking changes ^[5]. He is a longtime .NET developer who founded the continuous-integration provider AppVeyor in 2011 ^[3], and he started Flet in 2022 ^[2]. In packaged native applications, an in-process bridge has removed sockets from the channel between Python and Dart ^[14].

Flet says changes to its interface-diffing system produced gains of up to 6.7x in its own benchmarks ^[13]. For that number to reach your application, the diff has to be your bottleneck: a large control tree, frequent partial updates, and a device where the Flutter renderer Flet draws through ^[7] and the Python-to-Dart hop are not already the limit. Startup is measured separately. In a July 5th GitHub issue, one developer reported cold starts of roughly 10 to 13 seconds for an Android application built with Flet 0.85.3, covering a single application and predating the final release ^[15].

The published evidence of use is open-source reach. The repository had about 16,800 GitHub stars and 692 forks at publication ^[16], and Flet reports more than 9 million PyPI downloads, 100 contributors, 1,000 merged pull requests and 6,000 Discord members ^[17], which averages roughly ten merged pull requests per contributor ^[21]. Flet has not published revenue, paying-customer counts or the number of applications distributed through app stores ^[18].

Deployment shape still differs by target. On the web, Flet can run Python in the browser through Pyodide and WebAssembly, or keep Python on a server and send interface updates to the client ^[9].

In my view the stable label is worth taking at face value for new projects, because Fitsner took the breaking changes before he attached it. The cost sits with the people who wrote against 0.28. I would start new work on 1.0 and schedule the port of an existing application for a quarter when someone can own it.