AI delivery reliability

Your first question now gets a fast, real answer, PCs use the model they already downloaded, and response budgets match real hardware.

• Your first question now gets a fast, real answer instead of a timeout. The private AI assistant warms up in the background right after you launch the app, so by the time you ask, the answer comes back at full speed. Before this release the first question often paid the whole cold-start cost and could time out before it answered. If the assistant is still warming up when you ask, it now tells you so instead of quietly dropping to basic answers
• PCs that had already downloaded an AI model but were quietly falling back to basic answers now use the model they have. A memory check could previously steer the app toward a smaller model that was never downloaded, so it found nothing and dropped to basic answers even though a working model was sitting on disk. It now uses the largest model actually present that fits in your available memory
• Response time budgets are now tuned to real hardware. Instead of one flat time limit for every PC, the app sets its thinking time from what your class of machine actually measures, so a capable PC is not cut short and a modest one is not pushed past what it can do