• Constellation: use your Mac's models from your iPhone, iPad, or Vision Pro. Any Mac on your iCloud account with Remote Access turned on is found on its own — no address or token to enter — and tapping Use opens the chat right away while the Mac loads in the background. • Noema picks the most private way to reach your Mac for each message, from your local network through to an encrypted relay, and shows which one it is using in the chat header. Off-grid Mode now holds across every kind of connection, not just ordinary web traffic. • Your Mac can stand in for the cloud: its models appear in Autopilot's Stronger Model picker, and a new Fallback Model means a hard question tries another model before your on-device one answers. • Chats sync across your Apple devices if you want them to — one switch, off by default, in a new Sync & Devices setting. Conversations, bookmarks, and per-chat settings travel through your private iCloud account, and a reply that finishes on your Mac appears on your phone without opening the app. • Projects group related chats under shared instructions and sources, so every conversation in a project starts from the same context. Import PDFs, EPUBs, and text files straight into a project, or attach datasets you already have. • A rebuilt chat drawer brings search, projects, bookmarks, favorites, and date groups into one list with a preview of the last reply on every row, and chats stay noticeably faster during long generations. • Images you attach now reach the model at full resolution instead of being shrunk on the way in, so screenshots, receipts, and dense charts are legible to it for the first time. • Paged Overfit models run with the context you choose instead of a hidden cap, and Laguna S 2.1 now works as a Paged model. • Models are more honest about what they can do: a vision model that cannot actually read images no longer claims it can, and downloaded models use the Chat Template they shipped with. • On Mac, the Relay page is replaced by a simpler Developer page, and Remote Access has moved into Settings, under Sync & Devices. • Interface fixes, bug fixes, and polish throughout.

• Constellation: use your Mac's models from your iPhone, iPad, or Vision Pro. Any Mac on your iCloud account with Remote Access turned on is found on its own — no address or token to enter — and tapping Use opens the chat right away while the Mac loads in the background. • Noema picks the most private way to reach your Mac for each message, from your local network through to an encrypted relay, and shows which one it is using in the chat header. Off-grid Mode now holds across every kind of connection, not just ordinary web traffic. • Your Mac can stand in for the cloud: its models appear in Autopilot's Stronger Model picker, and a new Fallback Model means a hard question tries another model before your on-device one answers. • Chats sync across your Apple devices if you want them to — one switch, off by default, in a new Sync & Devices setting. Conversations, bookmarks, and per-chat settings travel through your private iCloud account, and a reply that finishes on your Mac appears on your phone without opening the app. • Projects group related chats under shared instructions and sources, so every conversation in a project starts from the same context. Import PDFs, EPUBs, and text files straight into a project, or attach datasets you already have. • A rebuilt chat drawer brings search, projects, bookmarks, favorites, and date groups into one list with a preview of the last reply on every row, and chats stay noticeably faster during long generations. • Images you attach now reach the model at full resolution instead of being shrunk on the way in, so screenshots, receipts, and dense charts are legible to it for the first time. • Paged Overfit models run with the context you choose instead of a hidden cap, and Laguna S 2.1 now works as a Paged model. • Models are more honest about what they can do: a vision model that cannot actually read images no longer claims it can, and downloaded models use the Chat Template they shipped with. • On Mac, the Relay page is replaced by a simpler Developer page, and Remote Access has moved into Settings, under Sync & Devices. • Interface fixes, bug fixes, and polish throughout.

• Noema Overfit (experimental): run models far larger than your device's memory would normally allow — even 100B-class models on a laptop — by streaming their experts from storage on demand. This is early and evolving, so expect rough edges. • PDF attachment parity: attach and read PDFs the same way across iPhone, iPad, and Mac, with consistent inline indexing everywhere. • Longer conversations that don't stall: new context compaction keeps chats going smoothly as they grow, so Noema stays responsive deep into a thread. • Fixed a model that could still appear loaded after it had actually been unloaded — for example when the app closed or the model was ejected unintentionally. • Fixed the key/value cache occasionally disappearing during a session. • Interface fixes, bug fixes, and polish throughout.

• Attach documents to chat on iPhone and iPad: add PDFs, EPUBs, and text files from the Files app and watch them index inline, so answers can draw on them right away. • Redesigned chat status drawer: compact capability controls, reasoning and Autopilot status, MCP visibility, and a clearer, more detailed context-budget meter. • Steadier iCloud sync across your devices. • Tidier model settings: a faster GGUF screen, with vision options shown only for models that support them. • Sharper math rendering for numbers and complex formulas, plus cleaner web citations. • Fixed a model that could still appear loaded after it had been unloaded. • Interface polish, bug fixes, and localization across every supported language.

• Connect to MCP servers: browse and add remote Model Context Protocol servers over HTTPS to give your models new tools, right inside chat. • Better PDFs in chat: models read through your documents more thoroughly and find the right information faster. • Smoother chat turns with cloud models — steadier, more reliable back-and-forth. • Make fuller use of your datasets straight from chat, so answers stay grounded in your own material. • Web Search improvements: models can now open pages and locate the evidence behind an answer. • More model settings to fine-tune how each model runs. • Interface fixes, bug fixes, and polish throughout.

• Noema Autopilot: switch it on and Noema sizes up each message, then routes it to the right model — local or remote — by difficulty. You get the best answer while saving energy and cost. • Study flashcards, made from your own material: turn any document or chat into a deck, then review it with spaced repetition to make it stick. • New reasoning toggle: turn a model's step-by-step thinking on or off, right from chat. • Reworked caching makes follow-up answers in document chats up to 2× faster — and up to 3–4× faster on large documents. • A refreshed Mac interface — cleaner, more responsive, and more at home on macOS. • More performance, stability, and interface polish throughout.

• Introducing Noema 2B — our own on-device model, now in Explore. It is tuned for noticeably stronger math, instruction following, and coding than its Qwen3.5 2B base, while staying small enough to run smoothly on your device. • New Tools tab: benchmark models head-to-head for speed and quality, and scan boarding passes to set up travel details automatically. • Hands-free Voice Mode (beta): speak to Noema and hear it answer, entirely on-device. • Logit & Jacobian Space viewer (Mac only): look inside the model's residual stream as it generates, then steer or swap direction vectors to shape its output — on-device interpretability you can actually touch. • Model pages now show the full Hugging Face README, rendered with clean, modern formatting and collapsible so it never takes over the screen. • Explore gains a Trending This Week section, surfacing the models people are downloading right now. • New curated Knowledge Packs in the dataset Explore page — offline, ready-to-search reference collections. • Smoother animations and a smoother search bar across the app. • Fixed the reasoning panel sometimes showing "interrupted" after a model had finished thinking normally. • More performance, stability, and interface polish throughout.

• New boarding pass scanning helps Noema detect travel details and configure passes automatically. • Improved context handling makes long conversations more reliable, with better rolling windows and smarter truncation. • Smarter dataset retrieval now follows your selected model settings more accurately. • Chat stays more responsive while models answer, files download, or datasets are indexed. • Chat has been cleaned up with fewer unnecessary controls, improved message actions, and better layouts for long model names. • Siri App Intents support lets Noema connect more naturally with system-level actions. • Voice transcription has been improved with clearer status handling and better reliability. • Apple-native model runtime support has been expanded, with better loading, downloads, and caching. • Explore has been refreshed with a cleaner catalog for discovering models. • Includes fixes for storage reporting, proxy support, memory settings, model loading, downloads, citations, and overall interface polish.

• Updated llama.cpp for Gemma 4 support, including fixes for previously known Gemma 4 issues • Improved GGUF import reliability, with better detection for chat templates, JSON configs, and multimodal projector files • Added a clearer download experience for CML models, including visible progress during downloads • Improved smart retrieval so large-context models make better use of available context with PDFs and long documents • Added a new Prompt Processing card in chat with live progress feedback • Fixed prompt processing progress getting stuck at 0% and corrected its placement after tool calls • Fixed scrolling issues in Model Settings caused by repeated memory fit checks • Updated VRAM estimates and maximum context recommendations to reflect KV cache quantization changes • Added support for Memory and system prompt customization • Refreshed curated models with Gemma 4 and Qwen 3 1.7B support

What’s New This update brings major improvements to tooling, model support, search, and overall app compatibility. New in this version: • Added support for the Python tool • Added OpenRouter support • Improved Explore recommendations for better model discovery • Updated LM Studio compatibility to REST v1, including a smoother model download flow through Remote Endpoint and Explore • Added ExecuTorch, CoreML, and Apple Foundation Model support • Updated the chat UI for a cleaner and more consistent experience • Updated llama.cpp and MLX to newer versions, including support for Qwen 3.5 • Fixed many bugs and stability issues • Removed SLMs

Release 2.0 brings native support for MacOS and VisionOS, introduces Noema Relay for seamless iPhone-to-Mac connection via CloudKit, adds vision support for models, enhances tool calling capabilities, rolls out localization across 10 languages, and ships accessibility fixes.

Noema is now completely free—unlimited web search and all features are included. FlashAttention and V-cache quantization cut memory use so larger models fit on more devices. You can benchmark any model or optimization to compare prompt-in/s and token/s. Input fields and remote endpoint forms are redesigned for clarity, and our llama.cpp core is updated to the latest upstream.

Noema 1.3 adds remote endpoints so you can connect OpenAI API, LM Studio, and Ollama servers over HTTP alongside your on-device models. New endpoint views surface provider details—architecture, quantization, context limits, and availability—while a guided setup handles registration, auth, and catalog refresh with smart warnings. Under the hood, bug fixes and reliability improvements keep hybrid chats fast and stable across local and remote models.

Release 1.2 brings reliability tune-ups and chat telemetry upgrades, keeping Noema responsive and insightful during conversations while preparing for upcoming tool automation. Includes bug fixes, stability improvements, better token counting, and faster chat speed stats. Also lays the groundwork for a new tool call coming soon.

Release 1.1 enhances Noema with important stability and compatibility improvements, ensuring sharper on-device reliability and readiness for the latest hardware. This update fixes key issues, including a bug with MLX model tool calling and a GGUF model loading bug on older devices. It also introduces new features: support for the iPhone 17 lineup, an updated llama.cpp framework, a friendlier user interface, and a more in-depth onboarding process designed to guide beginners. Full changelog: noemaai.com/changelog