Private, by default.
Conversations, attachments, settings, logs, and model metadata stay on-device. No account, API key, or remote inference service required. Optional web search is always opt-in.
Your data stays here
Intelligence, kept close
Your thoughts.
On device.
Pebble Mind is the private AI workspace for people who want to run Apple and Hugging Face models directly on iPhone. Powerful models, pro-level controls, and a conversation that stays yours.
Hugging Face → iPhone · No account · No API key · No telemetry
CHATSPebble Mind
⌕ Search conversations
L☷
Latest iPhone Model AvailableLlama 3.2 1B Instruct · 11:16 PM
The latest iPhone model available is the iPhone 17 Pro Max.
S☷
The benefits of a daily workout include…SmolVLM 256M Instruct · 11:12 PM
Daily workouts are beneficial for weight loss, strength and flexibility…
A☷
Explain why the sky is blue?Apple On-Device · 11:10 PM
The sky on Mars would appear reddish-orange. This is because…
◉ Private by default
⌁ Local inference
01 / The quiet advantage
Your models.
Your device.
For the first time, developers can bring supported Hugging Face GGUF models to an iPhone and work with them directly—without routing prompts through a hosted inference API.
It is private by architecture, not by promise. Download the model once, then chat locally.
01Hugging Face on iPhone
Validate and download supported GGUF models, then run them locally through llama.cpp.
02One calm workspace
Chat, model browsing, files, search, and settings share one native rhythm.
03Pro-level control
Tune context, reply length, temperature, top-K, top-P, KV cache, and quantization details.
02 / A considered workflow
From curious
to clear.
Pick a model that fits your device. Ask anything. Keep the trail. Pebble Mind turns the complex parts of local AI into a simple, tactile flow.
01
Choose your
kind of mind.
Browse text and multimodal models with readable size, memory, and readiness details before you download.
03 / The details matter
Built for the
way you think.
Native behaviors. Thoughtful feedback. The small things that make a powerful tool feel effortless.
Chat without friction.
Streaming replies, selectable Markdown, attachments, thinking cards, and a composer that is ready when the model is.
YouSummarize this locally● ● ●
Models, made legible.
See what fits, what is ready, and what it can do. Download only when you mean to.
72% of device memory · comfortable
Made for iPhone and iPad.
Push navigation, split views, native sharing, sheets, Liquid Glass, and accessibility in the places you expect.
iPhoneiPad
04 / Choose your depth
Free to start.
Pro when ready.
Every plan gets the private core. Pro is a one-time unlock for developers, power users, and anyone who wants more control over local models.
FREE
A private place to think.
Everything you need to experience local AI on Apple platforms.
- Local chat with Apple Foundation Models
- Browse and download supported bundled models
- Conversation history and local search
- Text and vision attachments where supported
- Optional web search, always opt-in
- Markdown replies, streaming, stop and retry
- Model info, readiness, memory, and download states
- App and model logs, plus reset controls
PRO · ONE-TIME PURCHASE
Your local AI lab.
Unlock the full Pebble Mind toolkit—built for developers and deep work.
- Bring your own models — add validated Hugging Face GGUF models by HTTPS, plus vision projectors
- Keep more models ready — manage multiple downloaded models
- Advanced generation controls — temperature, top-K, top-P, context window, reply length
- Enhanced limits — up to 65,536 context and 8,192 reply tokens where device-safe
- Performance details — generation metrics for completed replies
- Developer visibility — quantization info, KV cache type, searchable logs, log deletion
- Context compaction —
/compactcompresses long chats on-device while keeping the visible history - Transfer tools — export/import chats and configuration as local files
- Search domains — manage HTTPS web-search providers and allowlists
05 / When you want more
Meet
Pebble Mind Pro.
One thoughtful upgrade for people who want to go deeper with local AI. No recurring subscription.
Unlock the full experience ↗
ONE-TIME
PURCHASE
PURCHASE
Bring your own modelsAdd supported GGUF models from Hugging Face with optional vision projectors.
More room to exploreKeep multiple models ready and tune advanced generation settings.
Developer visibilitySee metrics, quantization, KV cache details, and searchable logs.
Your data toolsExport and import chats and configuration as local files.
Stay in the flowUse
/compact to focus long conversations without sending context away.05 / Good to know
The short
answers.
Does Pebble Mind send my chats to a server?+
No. Inference, conversation history, attachments, settings, and model metadata stay on-device. Optional web search is the explicit exception, and it is always your choice.
What models can I use?+
Use Apple Foundation Models where supported, plus bundled llama.cpp GGUF models. Pebble Mind Pro also lets you add validated custom models through HTTPS.
Is Pebble Mind Pro a subscription?+
No. Pro is a one-time, non-consumable purchase processed by Apple through StoreKit.
Which devices are supported?+
Pebble Mind is designed for iPhone and iPad, with model readiness and device-fit signals surfaced in the app.
Make space for better thinking
Ready when
you are.
Pebble Mind is available for iPhone and iPad with privacy at the center.