Cline Updates Cloud Session Features and Model Options

What happened

Cline v4.1.26 introduces new features for managing cloud sessions, including the ability to toggle between local and cloud models in API settings. Users can now select specific models for cloud tasks, which are stored separately from local models.

The update also includes a progress card that displays repository, branch, and steps when starting or resuming a cloud sandbox. Additionally, cloud sessions no longer show the Plan/Act toggle or the auto-approve bar, as these are now disabled by default.

Why it matters

These changes improve user control over cloud-based AI tasks and reduce potential confusion during session management. By separating model configurations for local and cloud environments, users can better tailor their workflows to specific use cases.

Key details

The model catalog has been refreshed, adding providers like CommandCode and Liquid Inference while removing others such as Llama. ClinePass now defaults to the Kimi K3.1 Pro model and includes additional models like MiniMax M3.1 Flash.

Model requests now retry a maximum of three times instead of five, and the Kimi K3 model supports the 'max' reasoning setting. GPT-6 models now use the 'max_completion_tokens' parameter instead of 'max_tokens'.