Cline Updates Model Defaults and Fixes Technical Issues
Cline introduces cloud sessions and adjusts model defaults in its latest release.

What happened
Cline has rolled out a new update that includes cloud sessions, allowing users to run tasks in a hosted sandbox using a GitHub repo and branch of their choice. The release also updates model defaults and addresses several technical issues.
The default model for Cline when no selection is made is now Gemini 4 Argon, which is the first model in the recommended list. Other providers, such as OpenRouter, continue to use their respective default models, like Claude Sonnet 5.5.
Why it matters
The changes to model defaults and catalog entries may impact users who do not specify particular models, as the defaults are now aligned with newer versions. Additionally, the technical fixes, such as handling long requests to local models and ensuring correct handling of OpenAI Compatible URLs, improve the reliability and user experience.
Key details
The update includes several technical adjustments, such as ensuring that images and files attached to answers are passed to the model instead of being discarded. It also fixes issues with long requests to local models like LM Studio and Ollama, which no longer result in a 'BodyTimeoutError' after 5 minutes of silence if the server is still processing.
Cline will retry a provider error up to 3 times before failing the turn if a response is interrupted mid-stream. Additionally, the model catalog has been expanded to include new providers, with some using MiniMax M3 as their default model.
Sources
Related stories

Gemini CLI Release Notes Highlight Changelog and Fixes
The latest Gemini CLI release includes a full changelog and several bug fixes.

Cline SDK Update Adds OpenAI Compatibility and Model Enhancements
Cline's latest SDK update introduces support for OpenAI-compatible GPT-6 models and expands the model catalog.

Meta Launches Forum App With AI-Driven Engineering
Meta introduced a new app called Forum, built using AI agents in the development process.

Krea Unveils Faster Image Editing Model
Krea's new image editing model is 4× faster than Nano Banana Pro, according to the company.

GitHub Introduces Evergreen for Secure Access and Repository Hygiene
GitHub's Evergreen workflow automates organization hygiene and secure access reviews.

Microsoft Introduces Magnetar for Secure, Fast Program Verification
Microsoft Research unveils Magnetar, a new method using zero-knowledge proofs to verify program execution efficiently.