What’s New in Version 2.0? LAPIS 2.0 is a major upgrade, rebuilt around Apple’s MLX framework to deliver faster, more efficient, and more reliable on-device AI. What’s new: - Native MLX inference across supported language and vision models. - Improved model loading and unloading with safer memory management. - More reliable vision conversations and image processing. - Redesigned model downloads with accurate progress, speed, and status information. - A greatly improved Live Voice experience powered by Apple’s native speech technologies. - Clearer voice output, faster responses, and smoother conversation flow. - A cleaner, calmer Live Voice interface with refined orb animations. - A more consistent and polished interface across chat, model management, and settings. - Improved animations, transitions, typography, and haptic feedback. - Better handling of long conversations and multimodal prompts. - Extensive stability improvements to prevent crashes during model loading and inference. - Internal architectural improvements that make LAPIS easier to maintain and expand. LAPIS 2.0 feels faster, more dependable, and more at home on iPhone and iPad—while keeping your AI experience private and on your device.

What’s New in Version 1.4 The overall experience has been refined so the interface feels smoother, more familiar, and more responsive. Live Voice Mode has taken a big step forward (clearer audio, faster reactions, better flow) still in preview, but now genuinely worth trying. You’ll also notice improved animations and more precise haptic feedback throughout the app. Finally, an issue affecting certain models—where responses degraded into “…” after several interactions—has been fixed.

What’s New in Version 1.3 This update brings upgraded voice output in Voice Mode (still in preview) with clearer, more natural voices. We’ve also added new recommended models that are lighter and faster to run on-device. Several user-reported bugs have been fixed, load times are shorter, and the overall experience should feel noticeably smoother.

What’s New in Version 1.2 A faster and smarter update focused on efficiency. This release introduces a new default LLM model that’s significantly smaller, faster, and more resource-efficient. Downloads and updates now complete much quicker, and token inference speed has improved noticeably across the app. Default parameters have been fine-tuned for better performance, smoother interaction, and reduced device load. Several user-reported bugs have also been fixed to keep everything running clean. Designed to give you a better offline, private, and on-device AI chat experience—now with less weight and more power.

What’s New in Version 1.1 We’ve listened carefully to your feedback and refined the Lapis experience. This update includes minor bug fixes and performance improvements across the app. On Mac and iPad, you can now send messages by pressing Enter, making conversations smoother and faster. We’ve also optimized performance on devices with lower resources—so larger models should now run more efficiently and with improved stability.