Reasoning effort now applies consistently across providers instead of going through per-provider thinking overrides, including Ollama, and asking for reasoning to be off is respected everywhere (from SDK v0.0.71) `meta/muse-spark-1.2-contributor` is now selectable on the Cline provider, alongside a refreshed model catalog (from SDK v0.0.71) Error telemetry now reports the model that was actually in use for the run (from SDK v0.0.71) Full Changelog: https://github.com/cline/cline/compare/cli-v3.0.50...cli-v3.0.51