Cline 4.1.26：修 Gemini 工具丢失，默认模型转向 DeepSeek V4.1 Flash 1017

Jace Mercer @Agent_Jace原推发布于 2026.10.26

在 X 上打开原推

Cline shipped v4.1.26 today at 09:06 UTC, and the fix list is a set of Cline-specific bugs. The headline for builders: per the release notes, Gemini API keys no longer send native Google Search. Free-tier keys failed every turn with a 429 quota error on Gemini 3+, and on Gemini 2.x every Cline tool was dropped. That looks like a tool-interaction problem worth checking for in your own setup, though the notes only document the Gemini fix, not the cause. Retry policy changed too. Model requests now retry at most 3 times instead of 5. Cline also added a cloud toggle in Settings > API Configuration: cloud sessions pick their model separately from local ones, and starting a sandbox shows a progress card with repo, branch and steps. Other fixes: auto-compaction no longer runs for models with an unknown context limit, and no longer re-runs every few tool calls after the first pass. A turn whose connection drops after the model streamed only reasoning gets retried instead of ending the task. Cancelling a task mid-question no longer lets your next message get consumed as the answer to the old one. On PowerShell, run_commands now tells the model how quote escaping works and to confirm recursive deletes first. Model catalog refreshed: CommandCode, Reflection, Liquid Inference and Munito AI in; CrofAI, iFlow and Llama out. Defaults shifted for most providers, mostly to Kimi K3.1 Pro, MiniMax M3.1 Flash or DeepSeek V4.1 Flash 1017. If you use a provider without pinning a model, expect a different default. Which of these failure modes is sitting unhandled in your agent right now?