Private AI,running locally.
Run language models, search your files, and use tools on your Apple devices. Your models, files, and conversations stay with you.
Native on iPhone iPad Mac Vision Pro
- On deviceGeneration and retrieval run on your own hardware.
- No accountPersonal local use needs no Noema account.
- Off-grid ModeBlocks external HTTP and HTTPS whenever you choose.
- FreeEvery feature, with no subscription.
The product
One workspace across your devices.
Models, files, and tools in one native app. Start on your Mac and continue on iPhone, iPad, or Vision Pro.
Capabilities
Everything you need for local AI.
One client for models, documents, research, and tools, designed for Apple silicon and built to work without a connection.
- 01
Run local models
GGUF, MLX, Core ML, ExecuTorch, and Apple Foundation Models in one native client.
- 02
Work from your files
Read PDFs more thoroughly and ground answers in your own datasets with retrieval that stays on device.
- 03
Research the web
Search the live web when a task needs it, then open the sources and check every citation.
- 04
Use real tools
Run Python locally, chart results, and connect MCP tools when the task calls for them.
- 05
Mac projects and Agent
Give Noema a working folder, review proposed edits, and approve commands before they run.
- 06
Larger models with Overfit
Run compatible mixture-of-experts models whose complete files would not fit in memory.
The privacy model
Local by default. Online only when you choose.
Local models, documents, indexes, and conversations can remain on your device. Optional features say where they connect, and Off-grid Mode stops them all.
Read the network activity guide| Feature | Destination |
|---|---|
| Local models and retrieval | Your device (on device) |
| Python, memory, and charts | Your device (on device) |
| Web Search | Search service you use (optional) |
| Constellation | Your devices and private iCloud (optional) |
| Remote endpoints | Provider you configure (optional) |
Off-grid Mode blocks external HTTP and HTTPS traffic from Noema. Local generation, retrieval, and tools keep working.
Introducing Constellation
Continue on another device. Use models from your Mac.
Chats, projects, and tool history sync across your Apple devices. Select a model installed on your awake Mac and use it from your iPhone or iPad. The model stays on the Mac; only requests and replies are sent.
Explore ConstellationLocal Network · connected
Noema models
Noema 1.5 4B
Noema's latest open model comes in two reasoning styles. Deep thinks longer on hard problems; Fast answers in a fraction of the tokens and almost never gets stuck. Both are post-trained from Qwen3.5-4B.
- Parameters
- 4 billion
- Styles
- Deep · Fast
- Release
- Open weights
- Released
- October 2026
- 98%fewer doom loopsFast on a hard-prompt stress test
- 3.1×fewer tokensSame full-budget GPQA Diamond accuracy
- 0/100long conversations loopedBoth styles; Qwen3.5-4B looped in 51
- Measured against stock Qwen3.5-4B in the same harness. Loop figures come from a stress test of hard prompts.
Noema for organizations
Private AI, adapted to your business.
NoemaAI works with organizations as well as individuals. Deploy Noema across your team, connect company knowledge, and adapt models to your workflows, with access managed through Noema Teams.
- 01Noema Teams deploymentEnroll employees, set role policy, and manage devices from one console.Available
- 02Company knowledgeAnswer from approved internal documents with retrieval that stays on device.Available with Teams
- 03Custom model pilotsAdapt a model to your terminology and formats, measured on your task and hardware.Pilot
Run models and work with files on your device.
Free on iPhone, iPad, Mac, and Vision Pro. No subscription, no cloud dependency.