Fetching works in an existing chat - Askingagain fetches again: Repeating a request like "fetch this site and summarize the headlines" in a chat that already has an answer now goes back to the web, instead of repeating the earlier summary. - No answer from thin air: If the assistant states figures or quotes for a live-data question without actually looking anything up, the app asks it to fetch before answering. - Today's headlines means today: An earlier reply is treated as out of date, not as the current answer.

Cancel Actually Cancels - Stop a model while it loads: Cancel now interrupts loading a multi-gigabyte model, not just the reply that follows it. - Stop a long prompt: Cancelling during the "reading your prompt" stage takes effect immediately instead of waiting for it to finish. - Stop an image mid-draw: Image generation now aborts partway through and gives the memory straight back. Cancelling is treated as something you meant to do — no error alert, no Retry button. You Can See What It's Doing - Loading progress for big models: A large model now shows "Loading model… 42%" instead of "Thinking…" for minutes on end. More Room to Think - 1.5× more context on 8 GB iPhones and iPads: 8192 → 12288 tokens, from two memory corrections rather than optimism — enough for noticeably longer documents and conversations. Better Answers from the Same Models - Models now follow their own recipe: A model can carry the sampler settings its author recommends inside the model file, instead of every model sharing one profile. - Correct prompt formatting: The special tokens that structure a conversation are handled as tokens rather than as ordinary text, which is how the models were trained to see them. Image Generation, Corrected - Your photos keep their shape: Image-to-image preserves the subject's proportions instead of stretching everything to a square, and respects the orientation your phone recorded — portraits no longer come back rotated and cropped. - Reproducible images: The seed appears with the result, so you can recreate an image or vary it deliberately. - Smoother previews: Live previews no longer run a full decode on every single step. A Smaller, Newer Model Catalog - Gemma 4 E4B joins, alongside Gemma 4 E2B — and Gemma 4 is now Apache-2.0 licensed. - Older models retired: Gemma 3, Phi-4 mini, Ministral 3 8B and Mistral Nemo Prism are no longer offered to new installs. Anything you have already downloaded keeps working; retired models stay in your list until you delete them. - 12 text models offered, down from 16 — chosen so each one covers a device class the others do not. More Reliable Tools - Tool parameters from MCP servers now carry their descriptions and numeric limits through to the model, union-typed and combined-schema parameters are no longer dropped, and a tool that takes no arguments no longer has one invented for it. Privacy & Stability - Prompt-injection defences now derive their markers from the model's own vocabulary rather than a fixed list. - A build-time guard fails the release if the shipped image engine is not the exact version the build pins, and a data race in image-model downloads is fixed.

A big update: new image engine, a fresh design, and more models • Faster on-device images — A new Stable Diffusion engine runs fast image models like SDXL Lightning and Z-Image Turbo right on iPhone. Watch a live preview as it draws, and guide results with image-to-image, seeds, and negative prompts. • Browse your creations — A new Gallery collects every image you've made; open any one full-screen and jump back to its chat. • Redesigned everywhere — A fresh look across iPhone, iPad, and Mac with Dark Mode, first-run onboarding, and a roomy sidebar for grouping and pinning chats. • More models — Added Gemma, Qwen, Llama, Ministral, and Phi families. Set a different model or persona for each chat. • Smarter, snappier chat — More reliable tools on small models, richer Markdown, and faster first replies. • More document types — Add documents from a file or a URL, now including Word, RTF, and PDF. • Connect your tools — Add custom HTTP headers for MCP servers, stored securely in the Keychain. Thank you for trying Sessions AI. We hope it helps you get more done without giving up your privacy.

On-Device Image Generation - Fast image models on iPhone: A new in-process Stable Diffusion engine runs GGUF image models directly on iOS — no cloud, no APIs. SDXL Lightning generates in just a few steps for quick results. - A bigger model lineup: Choose from Stable Diffusion, SDXL Lightning, Z-Image Turbo, and FLUX.2 Klein, plus Qwen Image on high-memory Macs. A lightweight Stable Diffusion 1.5 option fits 4 GB-RAM devices. - Watch it draw: See live previews and phase-aware progress as an image is generated, with the generation time and model name shown right in the chat. - More creative control: Guide results with image-to-image, a seed for reproducibility, and negative prompts. - A real Gallery: Browse everything you've created in a cross-platform Gallery, open any image full-screen, and jump back to the chat that made it. - Respects your device: Image generation is offered only when it can fit in memory alongside your language model, and pauses when the app is backgrounded. Redesigned Across iPhone, iPad & Mac - A fresh look: A new design system with full system dark-mode support. - Built for each device: A floating tab bar on iPhone, and a grouped, pinnable sidebar with quick access to Models, Gallery, and Settings on iPad and Mac. - Easy first run: New onboarding welcomes you and helps you pick and download your first model. - Per-chat models and personas: Set the model and persona for each conversation, shown as chips while you compose. Settings are now organized as "Personas." - Better reading: Richer Markdown rendering for headings, lists, quotes, code, and inline formatting. - More control over chats: Edit a chat or delete an individual prompt and its reply. More Models & Faster Replies - New model families: Added Gemma, Qwen3 and Qwen3.5, Llama 3.2 and 3.3, Ministral, and Phi-4-mini, with clear labels for deprecated models. - Reliable tools on small models: Smaller models now call tools far more reliably, so features like calculator, search, calendar, and image generation work even on compact models. - Snappier responses: Models pre-warm when you open a chat, and conversation context is reused between turns for faster replies. Documents & Knowledge - Add documents from a URL: Pull in a web page or document by URL — for a single chat or for your persona's knowledge base — and review the extracted text before adding it. - Rich-text support: Attach and ground answers in Word (.docx), RTF/RTFD, and ODT documents, all parsed on device. Tools & Connections (MCP) - Custom headers for MCP servers: Add custom HTTP headers (such as API keys) when connecting to MCP servers; header values are stored securely in the Keychain. - Built-in tools always available: Connecting an MCP server no longer hides the app's built-in tools. - Better-grounded answers: Web search and page fetching pull in more content and stick to it, reducing made-up details, with more reliable handling of long results.

Performance Improvements - Faster Image Generation: Core ML models now use optimization hints that prioritize inference speed for Stable Diffusion models - macOS GPU Optimizations: Enabled FP16 accumulation on GPU for faster image generation on Mac devices Code Quality - Cleaner Code: Refactored trace reading and model configuration into reusable, well-documented methods

New Features: - Help Menu: Access documentation directly from the app via the new Help menu item Bug Fixes: - Fixed a crash that could occur on older iPad Pro models (3rd generation) during Metal operations - Fixed chat list search functionality Performance: - Improved memory management for image models and image generation based on your device's available RAM