Claude 5.5, from Haiku to Opus
Sonnet 5.5, Opus 5.5 and Haiku 5.5 now power Dopia, and every model picker gets an effort dial so you decide how hard the model thinks.
Claude Sonnet 5.5 and Opus 5.5
Anthropic's newest generation is in. Sonnet 5.5 is now the default for new chats, Opus 5.5 is there for the hardest questions, and anything you had set to an older Claude moved over by itself.
Claude Haiku 5.5
Haiku 5.5 joins the picker as the fast, low-cost Claude. It's also the new default for the work Dopia does in the background — deal memory, email drafts and routines — unless your workspace picks otherwise.
- →Picked by running each job on real data, where it beat the model it replaced on all three
An effort dial on every model
How hard a model thinks decides both the answer and the credits it spends. The dial you know from chat now sits beside every model picker — on each routine, and for each job under Organization settings → AI models — and each job starts at the level it tested best at.
- →Each model offers its own levels — five on Claude and GPT-5.6, three on DeepSeek and GLM
- →Max is marked for what it costs, four times the usage or more