September 22, 2026
Lovable's Build Process Chooses Its Own AI: Claude Opus 5.5 and GPT-5.6 Sol
On September 22, Lovable added Claude Opus 5.5 and GPT-5.6 Sol to automatic model selection. In the original post, the company called the latter model GPT-6 Sol, but OpenAI's documentation and Lovable's changelog call it GPT-5.6 Sol. Just describe the task: the platform handles the build itself and distributes its parts across models.
Lovable
@lovable
Two new models are available in Lovable today: Claude Opus 5.5 and GPT-6 Sol. You don't need to think about which one to choose. Lovable combines the best models so you get the result you need.
Pretty much every week, a new AI model drops. (Speaking of: Lovable now builds with Fable 5.1) We test each new model, tune it, and decide how it’s routed to ensure your builds have the best outcome. Our job is to keep up, your job is to keep building. Read more about our model philosophy: http://lovable.dev/blog/the-model-picker-is-a-dead-end
· 10.4K views
Previously, Lovable built with Fable 5.1. Now Lovable selects the model for the task itself and can change the route during the build.
Fewer steps. In Lovable's tests, Claude Opus 5.5 required 26% to 48% fewer steps for a new build and 34% to 47% fewer for refining code. On the result-verification metric, it outperformed Opus 5 by 4-6% and used 42% to 57% fewer steps.
GPT-5.6 Sol supports 1 050 000 tokens of context and can produce up to 128 000 tokens per response. For AI features inside an app, the model is connected by asking Lovable, without your own OpenAI key; usage is calculated by tokens.
Lovable plans to test new models and change routing as they are released.
Source
