KernelAI 2.0 is the biggest update we have ever shipped, and it did not come from a roadmap, it came from you. Every piece of feedback, every feature request, every "please just add this" got read, and this release is the answer. We rebuilt the app to feel native from the ground up: a new navigation drawer, smoother streaming, cleaner chat, and dozens of fixes you will feel the moment you open it, even before you notice a single new feature. What is actually new: • Chat with PDFs, DOCX, CSVs, Markdown, and text files on-device • Ask questions about photos, screenshots, charts, and diagrams with vision models • Choose from 50+ device-matched models, including Qwen 3.5 and Gemma 4 • View generated thinking output from supported reasoning models • Dictate prompts with voice input • Organize work with folders, search, branching, custom prompts, and export • Ask KernelAI from Siri and Shortcuts on supported iOS 26 devices • Reliability, performance, and accessibility improvements throughout Download a model once, then test local chat in Airplane Mode. No account or subscription required.

KernelAI 2.0 is the biggest update we have ever shipped, and it did not come from a roadmap, it came from you. Every piece of feedback, every feature request, every "please just add this" got read, and this release is the answer. We rebuilt the app to feel native from the ground up: a new navigation drawer, smoother streaming, cleaner chat, and dozens of fixes you will feel the moment you open it, even before you notice a single new feature. What is actually new: • Chat with PDFs, DOCX, CSVs, Markdown, and text files on-device • Ask questions about photos, screenshots, charts, and diagrams with vision models • Choose from 50+ device-matched models, including Qwen 3.5 and Gemma 4 • View generated thinking output from supported reasoning models • Dictate prompts with voice input • Organize work with folders, search, branching, custom prompts, and export • Ask KernelAI from Siri and Shortcuts on supported iOS 26 devices • Reliability, performance, and accessibility improvements throughout Download a model once, then test local chat in Airplane Mode. No account or subscription required.

- New: Qwen 3.5 models (4B, 2B, 0.8B) — the latest hybrid architecture with built-in reasoning and support for 201 languages - Upgraded inference engine for faster and broader model compatibility - Fixed thinking mode display for cleaner AI responses

Translate text in 50+ languages — entirely on your device. • New translation mode powered by TranslateGemma • Native language picker with popular languages pinned for quick access • Swap languages instantly with one tap • Copy translations to your clipboard • Improved prompt suggestions • Bug fixes and performance improvements

• Listen to AI responses - Tap the speaker icon to have messages read aloud • Smarter model recommendations - We now suggest models optimized for your device • Better memory management - Models that require more RAM than your device has are now clearly marked

- Improved stability - Fixed crashes that could occur when loading AI models on some devices - Better error messages - Clearer feedback when downloads fail, with helpful suggestions - Recovery options - If something goes wrong, you can now easily switch models or reset the app - Bug fixes - Various improvements for a smoother experience