What’s New This release introduces a new generation of Playground for LM, bringing local intelligence, multimodal understanding, and smarter model workflows together in one unified experience. On supported devices, Playground for LM can now run compatible MLX models locally, taking advantage of Apple silicon to deliver faster, more private on-device inference. Multimodal workflows have also been redesigned. Native multimodal models can understand images directly, while text-based models can now use a VLM-powered fallback to analyze visual content and convert it into useful text context. This makes image understanding available across more models, even when native vision support is not built in. This update also adds beta support for RAG, allowing models to reference external knowledge and generate more context-aware responses. A smarter routing system now helps choose the right path across local models, vision models, multimodal models, retrieval workflows, and API-based execution. For advanced workflows, the new beta API makes it possible to connect Playground for LM with external tools and use available compute resources, including computer-side processing when configured. These features are currently in beta and may affect stability in some situations. Compatibility, performance, and behavior may vary depending on your device, model, and setup.

What’s New in Version 1.8 Version 1.8 is one of the most important updates yet for Playground for LM — a full-scale evolution of local AI performance, stability, speech workflows, and interface polish across Apple devices. At the core of this release is a carefully rebuilt architecture, designed to better adapt to a wide range of hardware configurations. From iPhone and iPad to Mac and Apple Vision Pro, Playground for LM now coordinates model loading, memory usage, and task switching more intelligently, bringing a smoother and more reliable on-device AI experience to every supported device. Local intelligence is now significantly more stable under high-load scenarios. With a new model scheduling strategy, more precise memory coordination, and more resilient recovery logic, stability can improve by up to approximately 90% in supported workflows — without compromising performance. Speech and transcription have also been upgraded throughout. Whisper-based voice input and audio transcription are now smoother, faster, and more reliable, with intelligent progress recovery that allows interrupted transcription tasks to continue from a saved point instead of starting over. Supported transcription history items can now automatically generate clearer, more readable titles, making long-term content management easier than ever. Language models and speech models now work together more intelligently. When switching between local chat, voice input, audio transcription, and model recovery, Playground for LM can better allocate resources, reduce unnecessary model unloads, and recover more gracefully when temporary memory pressure occurs. Version 1.8 also introduces a new Chip Performance page, featuring Apple chip badges and clearer capability explanations. Users can now better understand how their device supports model loading, speech recognition, multitasking, and continuous on-device AI workflows. The interface has been refined across the app, including long-message reading, Markdown rendering, table and code previews, waveform behavior, return-position handling, colorful intelligent loading animations, completion states, sound feedback, haptics, Liquid Glass visual details, theme consistency, localization coverage, and offline legal document fallback. This release also includes extensive reliability fixes and refinements for scrolling, model recovery, transcription history, title animations, large-screen input behavior, and overall stability. Built for local AI. Ready for higher loads. Designed to recover faster, run longer, and feel more refined on every Apple device.

What’s New in Version 1.7.5 With version 1.7.5, Playground for LM becomes faster, smoother, and more deeply refined for the way you work with local intelligence. This release brings improvements across performance, visual feedback, model orchestration, and rich content presentation. Together, they make every conversation feel more responsive, more capable, and more naturally connected to your workflow. A smoother foundation We’ve optimized the app from the ground up to improve speed, stability, and overall efficiency. Conversations now feel more fluid as you scroll, write, and move between different types of content. Message bubbles render more reliably, complex layouts feel more consistent, and the interface responds with a clearer sense of rhythm across devices. Every detail has been tuned to make Playground for LM feel lighter, steadier, and more effortless in everyday use. Intelligence, more deeply integrated Version 1.7.5 also improves how the app works with local AI models. Model activity, context changes, and interface feedback now come together in a more coordinated way. Instead of feeling like a separate process running behind the conversation, intelligence now feels more naturally woven into the app itself. As you write, the app prepares context. As the model works, the interface responds. And as results appear, the content is presented with greater clarity and structure. It creates a more seamless local AI experience — one that feels immediate, understandable, and thoughtfully integrated from start to finish. Visual feedback with more expression We’ve refined the way Playground for LM communicates intelligent activity. Loading, processing, and completion states now use richer and more precise visual feedback, helping you understand what’s happening without pulling attention away from the conversation. The interface feels more alive when intelligence is at work, while remaining clean, focused, and familiar. It gives each moment of waiting a clearer purpose, and each response a stronger sense of arrival. Rich content, right in the conversation Structured content is now easier to recognize, preview, and work with. When you send tables or code blocks, message bubbles can display elegant previews directly in the chat. This makes it easier to understand what a message contains at a glance, whether it’s a code snippet, a data table, model settings, or organized notes. Conversations can now carry more than plain text. They can hold the structured, technical, and creative content that real work often depends on — without making the chat feel cluttered. A better space for longer messages Longer messages now have a more comfortable place to live. When a message becomes too large for a standard chat bubble, it can open in a dedicated reader view designed for focused reading. This keeps the conversation clean while giving long-form content the space it deserves. The reader supports full rendering across content types, including Markdown, rich text, tables, code blocks, and advanced layouts. Whether you’re reviewing a long prompt, a detailed answer, a document-style message, or a complex piece of structured content, everything is easier to read, understand, and navigate. Built for a more complete local AI workflow Playground for LM is designed to make local AI feel powerful, approachable, and native to Apple devices. Version 1.7.5 continues that direction with a smoother foundation, smarter model interaction, more expressive visual feedback, and better support for rich, complex content. It makes the app feel faster. It makes conversations feel clearer. And it makes local intelligence feel more naturally woven into the way you create, explore, and think. This is Playground for LM 1.7.5 — more refined, more capable, and built to make every conversation feel beautifully considered.

Good news: Version 1.7 is now available. With Version 1.7, we’re bringing a more refined local AI experience to Playground for LM — one that feels more natural in conversation, more immediate in transcription, and more deeply integrated with the way you use your Apple devices every day. This release is about making local intelligence feel closer, faster, and more effortless. Whether you’re typing, speaking, transcribing, or listening, Version 1.7 introduces thoughtful improvements across the app to make every interaction feel more responsive, more polished, and more complete. Voice input is now built directly into the chat input bar. When a local Whisper model is already available on your device, Playground for LM makes it even more useful by bringing local speech recognition right into the conversation. You can speak naturally and turn your voice into text without leaving the chat. When the input field is empty, recognized speech can be sent automatically as a new message. And on larger layouts, including iPad and Mac compatibility environments, recognized speech can be appended directly to your existing draft, making it easier to continue an idea without starting over. We’ve also refined the recording experience to make it clearer and more expressive. A real-time waveform gives you immediate feedback while you speak, while transcription and loading status now appear in a centered, easier-to-read position. The result is a voice input experience that feels focused, understandable, and right at home inside the chat interface. Version 1.7 also introduces a built-in on-device intelligent title organization system. Playground for LM can understand conversation content locally and create clearer, more recognizable titles for your history items when the local model is available. It is designed to work quietly in the background while giving user-initiated model responses top priority, so your active conversations stay uninterrupted. Transcription history has been redesigned with native Liquid Glass-style cards, bringing a more modern, layered, and consistent look to saved transcription items. Conversation drawer spacing has also been improved, ensuring the final history item remains fully visible and easier to access. Audio playback now feels more integrated with the system. Version 1.7 adds Now Playing artwork that matches your selected app icon, paired with a clean black background for a more elegant Dynamic Island presentation. It’s a small detail, but one that makes playback feel more intentional and visually consistent across the system. This update also expands localization coverage across non-legal app content, making more of the experience feel complete in supported languages. The disclaimer and bundled offline legal document fallback content have also been updated for Version 1.7, helping ensure important information remains available even when the app is used fully offline. Under the hood, Version 1.7 includes stability and task-priority improvements designed to make the app feel more dependable during real use. User-initiated model responses now take precedence over background maintenance tasks, keeping the app focused on what you’re doing in the moment. Version 1.7 brings together smarter local organization, better voice input, improved transcription history, UI refinements, reliability improvements, and a more polished playback experience across iPhone, iPad, and Mac compatibility environments. It’s a more complete local AI experience — designed to feel thoughtful, capable, and distinctly at home on Apple devices. And now, back to the experience.

What’s New This update adds local audio transcription powered by Whisper.cpp and introduces a new limited free trial for Playground for LM. You can now try the core local AI workflow before upgrading to Playground For LM Pro. Free users can test local model chat with a limited message allowance and try local transcription with a limited allowance. Playground For LM Pro unlocks the full experience, including extended local model usage, full transcription access, premium themes, and the exclusive gilded Pro app icon. New in this version: • Local speech-to-text transcription powered by Whisper.cpp • Support for recorded and imported audio transcription workflows • On-device transcription for supported audio • Multilingual transcription support depending on the selected model and audio content • Transcript review and text organization improvements • New transcription status and runtime information • A limited free trial for core local AI and transcription workflows • Playground For LM Pro unlocks full access, premium themes, and the exclusive Pro icon • Improvements to the local AI model workspace • Stability and interface refinements Playground for LM continues to focus on private local AI workflows on Apple devices, including llama.cpp-based local model inference, streamed conversations, model management, clean content rendering, diagnostics, and now Whisper.cpp-powered audio transcription.

This update includes minor improvements and fixes for some common issues to make Playground for LM more stable and smoother to use.