• Gemma 4 E-series models (E2B/E4B) no longer leak internal reasoning text into their answers.
  • Qwen models no longer flip into thinking mode when they shouldn't.
  • Opening Settings no longer re-verifies installed model files, removing a noticeable stall on the Local AI settings screen.
Local AI runs on Apple Silicon Macs, Windows, and iPhone 15 Pro or newer.