Cline v4.1.23 Adds New Model Support, Improves Context Handling, and Updates Defaults
Cline v4.1.23 introduces improved multi-turn handling, expands accessible models, and updates default settings, while fixing handling for certain APIs and model quirks.
What changed?
Cline v4.1.23 brings several key updates. If a model’s response ends without a recognized finish reason, Cline now prompts the model to continue once instead of assuming completion. The model catalog was refreshed: GPT-6.1 Sol is now in the recommended list; Solar Mini 4 is added and DeepSeek V4.1 Flash and space-bunny-alpha are removed from the free list; multiple provider defaults are also updated. Vultr model IDs were changed upstream, requiring users with pinned Vultr models to re-select a model. API compatibility is improved, including fixes for using Claude with custom Anthropic URLs and for models like Kimi K3 that only accept certain reasoning levels. Additionally, when an MCP tool gives more output than fits in context, Cline now allows users to access the remaining output incrementally.
Why does it matter to an everyday developer?
These changes improve usability and reliability for developers working with Cline's broad set of AI models and tools. The improved handling of incomplete responses increases robustness for longer interactions or edge cases where models do not signal completion as expected. Expanded and updated model lists mean access to newer or better-performing models with less manual configuration. Fixes for custom API backends and model-specific quirks reduce failure points, saving developer time. The new context overflow solution for MCP tools ensures that outputs exceeding the context window are still accessible, which is critical for reviewing or acting on large outputs in workflow integrations.
What can the developer do now?
Developers using Cline should check if they have any pinned Vultr models and re-select them if necessary, as missing or changed IDs may affect their workflows. They can now try new models like GPT-6.1 Sol or Solar Mini 4, and review any updated defaults among providers to confirm desired behavior. For those integrating MCP tools, developers can rely on the tool to access overflow results via provided previews and links. Those using custom Anthropic endpoints or working with models enforcing strict reasoning levels should see improved reliability and can resume or expand those workflows confidently.
