Edge AI · On-Device LLM

The AI that runs
without the internet.

OceanAI's on-device LLM runs fully offline on your Android or iPhone. No server. No cloud. Your conversations with your health AI never leave your device unless you choose to escalate.

Download OceanAIAll features
Why it matters

Why on-device matters

100% private
Your health data never touches a server. Conversations, diagnoses, and lab results stay on your phone.
Works anywhere
No Wi-Fi, no mobile data, no problem. Full AI capability in a hospital, on a mountain, mid-flight.
Zero latency
Responses come from your own CPU/GPU — no round trip to a server. Instant.
No subscription risk
On-device AI doesn't go offline when a cloud provider has an outage.
Technical stack

What's actually running on your phone

Inference engineMediaPipe C++ native bridge
ModelGemma-2B quantized (INT4)
PlatformAndroid 12+ · iOS 16+
Memory footprint~1.2 GB RAM
FallbackCloud LLM when device limits reached
Architecture

Hybrid edge + cloud, by design

0ms
Network round-trip for on-device replies
2
Inference tiers — device, then cloud
INT4
Quantization for phone-class memory
100%
Health conversations private by default

Ocean AI dynamically escalates from the on-device model to a cloud endpoint only when a query needs complex clinical logic — full symptom analysis, insurance code mapping, or multi-document reasoning. Everyday questions, reminders, and quick lookups never leave the phone.

FAQ

Common questions

Why not always use the cloud model?

The cloud model (claude-sonnet-5, via OceanAI's own API route) is more capable for complex clinical reasoning — the on-device model handles routine questions instantly and privately, then hands off automatically when a query needs deeper reasoning.

Does offline mode work for file uploads too?

Simple text documents can be extracted fully on-device. Image and PDF extraction currently requires the cloud model, since vision inference isn't yet part of the on-device pipeline.

Will older phones support this?

The on-device model requires roughly 1.2GB of free RAM to load — supported on Android 12+ and iOS 16+ devices from the last several years. Older or lower-memory devices fall back to cloud inference automatically.

Explore more of the platform

📚

AI Health History

Every AI conversation stored, searchable, and contextual.

🩺

Doctor Appointments

Book and manage doctor-patient appointments inside the app.

🩸

Nearest Blood Donor

Real-time proximity matching for compatible blood donors.

Your AI, running in your pocket.

On-device inference ships in every build of OceanAI.

Download OceanAI