Cline CLI 3.0.65 compacts and retries instead of failing when local models hit the output cap, and moves 11 providers' default to Opus 5.5

GitHub·medium signalXBlueskyLinkedInCopy link

llama.cpp, Ollama and LM Studio cap generation at whatever context is left. Before Cline CLI v3.0.65, that cut off long local-model sessions mid-answer and failed the run. Now the CLI compacts the conversation and retries the turn once. The model catalog grows from 6,237 to 6,386 models across 209 providers, and the default changes for 19 providers. Eleven of those, including GitHub Copilot and Vertex, now default to Claude Opus 5.5, so users who never picked a model will silently get a different one.