QVAC

QVAC vs Cactus

QVAC

Open-source ecosystem for local-first peer-to-peer AI on every platform.

Cactus

A hybrid edge-cloud inference engine for phones and wearables.

Key differences

QVAC and Cactus are both built to run local AI on phones and wearables. Cactus ships its own kernels and quantisation rather than wrapping llama.cpp, publishes bindings for Swift, Kotlin, Flutter, React Native, Python and Rust, and covers text, vision, transcription, embeddings and RAG. QVAC runs on Fabric, a llama.cpp fork, and exposes JavaScript, TypeScript and Python.

The licences differ in scope. The Cactus engine is source-available: its free grant covers individuals, education, non-profits, and organisations with both under two million dollars in funding and under two million in annual revenue, and crossing either threshold calls for a commercial licence within thirty days. The cactus-react-native package separately declares MIT, so the engine LICENSE file is the governing document. The QVAC SDK is Apache 2.0 and Fabric is MIT, with no thresholds attached.

The architectures point in different directions. Cactus is hybrid edge-cloud, with cloud handoff as a designed path when a device cannot carry a request. QVAC keeps work on devices and can delegate to another device over an encrypted peer-to-peer link. Whether a request may leave the device decides which one fits.

This page compares Cactus v2.1.0, released 17 August 2026, against QVAC 0.18.2, meaning the SDK together with the Fabric inference engine at v10297.1.1. Every row was checked against the project's own documentation and release notes on 4 September 2026. Both projects move quickly, so check the current release before you make a decision on either one.

Feature matrix

Feature

QVAC

Cactus

PLATFORMS

macOS

Yes
Yes

Windows

Yes
No

Linux

Yes
No

Android

Yes
Yes

iOS

Yes
Yes

AI TASKS

Text generation

Yes
Yes

Transcription

Yes
Yes

Translation

Yes
No

Image generation

Yes
No

OCR

Yes
No

Text-to-speech

Yes
No

RUNTIME SUPPORT

Node.js

Yes
No

Bare

Yes
No

Expo

Yes

Via RN

HTTP server

Yes
Yes

CLI

Yes
Yes

P2P

Peer discovery

Yes
No

Inference delegation

Yes

Cloud handoff

Encrypted transport

Yes
No

MOBILE SUPPORT

On-device inference

Yes
Yes

LoRA fine-tuning on mobile

Yes

Merge only

Mobile SDK

Yes
Yes

LICENSING

License

Apache 2.0

Source-available

Open weights tooling

Yes
Yes

When to choose QVAC

Your organisation is funded, earning, or expects to be.

You need a permissive licence with no revenue threshold.

You need every request to stay on the device.

You need OCR, translation, image generation or LoRA fine-tuning.

When to choose Cactus

You qualify under their free tier and expect to stay there.

You are targeting wearables specifically.

You want Flutter, Swift, Kotlin or Rust bindings.

Cloud handoff for heavy requests suits your product.

Compare

Ready to build with QVAC?

One SDK, every platform, no rent. Grab it and ship your own local-first AI.

npm install @qvac/sdk