Cline swapped your default model

What changed

If Cline suddenly feels different, it may not be you: the default model changed.

What changed

This Cline release changed what happens when you never picked a model. With no model chosen, the Cline and OpenRouter providers now use Claude Sonnet 5.5.

The release also refreshed the model catalog: it adds Claude Haiku 5.5, and the default model changes for a long list of other providers, including GitHub Copilot, Google Vertex AI, Requesty and NanoGPT. Most of those now default to Claude Haiku 5.5 or Muse Wonder 1.0.

How it affects your work

A different model can answer differently, follow your instructions differently and cost a different amount per task. If you never pinned a model, your next session runs on the new default without asking you first.

What to do

• Open Cline’s settings and check which model is selected for your provider.
• Liked the old one? Pick it explicitly, so the next catalog refresh can’t swap it.
• After a few tasks, glance at your usage or cost dashboard.

Also in this release: turning reasoning off no longer fails with a 400 error on GPT-6 Astra, GPT-6.1 Sol, Claude Fable 5 or Claude Opus 5.5, and screenshots no longer inflate the token estimate, which had made auto-compaction run long before a conversation was near its limit.