On September 1, Anthropic released Fable 5.1, a new model designed for coding and long-running tasks. That same day, Lovable announced it was building applications using the model.
According to Lovable’s measurements, iterative code editing has become 17% more accurate and 31–39% cheaper, depending on the task's complexity. UI-related work saw a 3.5% improvement. There were no gains in building applications from scratch; the entire benefit came from refining apps that were already up and running.
That is where the bulk of the work lies. CTO and co-founder Fabian Hedin puts it this way: Fable 5.1 pulls furthest ahead of the previous model precisely in the area where users spend the most time—taking an already running application and teaching it to do more.
For the platform, this is primarily about cost efficiency: the most frequent user operation has become roughly one-third cheaper and more accurate.
On September 7, Lovable complemented this with "Drafts"—parallel project versions that can be built without affecting the live application. Together, these moves over the course of a week steer the product in a specific direction: shifting from generating an initial screen to an environment where a live product can be further developed.