PearPie Chat
Built for desktop and iPhone
Open it, pick a model, talk. PearPie Chat is where local models, premium models on the PearPie Network, and your linked devices all meet in one window.
PearPie Chat is a regular app, whether you install it on your computer or your iPhone. You open it, you're in. No sign-up, no setup wizard. The interface is a familiar chat window: conversations, a model picker, and an input area. Curious why we built it this way, or how the privacy holds up underneath? Read the about page for the why and how it works for the mechanism.
The interface
Simple by default. Advanced when you want it.
The same window holds your conversations and your synced devices, and lets you talk to both local models (downloaded onto your machine) and PearPie hosted models without switching interface or settings.
Simple mode works like the AI apps you already use. You don't have to think about whether a model is running on your device or on the PearPie Network. The app figures out the routing and you just type.
Advanced mode gives you full control of the peer-to-peer side: turn remote inference on or off, pick which devices you sync with, decide who you share credits with.
What's in the box
Local models, one click away
A built-in library of open-source models you can download and run directly on your hardware. Gemma 4 works on most machines (RAM is the main requirement), and Qwen 3.5 unlocks on machines with a GPU. More coming.
Premium models when you need them
Claude, DeepSeek, Mistral and others, served via the PearPie Network on European infrastructure. You pay only for what you use.
Linked devices, in sync
Pair a second device and your chats follow you. You can also borrow a stronger device's GPU from a smaller one. How that works →
No account, ever
A private key generated on your device is your identity. There's nothing to sign up for and nothing for us to lose.
What's available in this pilot
An honest snapshot of what's in the box right now:
- Platforms: macOS, Windows, Linux, and iPhone. All four launch together.
- Local models: Gemma 4 (runs on most hardware) and Qwen 3.5 (GPU required). More open-source options being added.
- Premium models (PearPie Network): Claude, DeepSeek, Mistral.
- Peer-to-peer device linking: Works between any combination of desktop and iPhone.
- Bring your own model server (Ollama): Coming soon.
If you join the pilot we'll let you know when each new piece lands. Feedback during this period genuinely shapes what gets built next.