← Capa | Mercado

SCX.ai bridges Australian sovereign LLM Project MAGPiE into Apple's Foundation Models framework

2026-11-04 · Fonte:ARN APAC

SCX.ai bridges Australian sovereign LLM Project MAGPiE into Apple's Foundation Models framework

MIT-licenced Swift package SCXSwiftKit lets Apple developers route sessions to locally hosted Australian model Project MAGPiE while retaining native LanguageModelSession, @Generable, tool and streaming APIs.

Australian AI firm SCX.ai is building a bridge between Apple's Foundation Models framework and externally hosted large language models, with its locally served sovereign model Project MAGPiE positioned as the flagship destination. The effort centres on SCXSwiftKit, an MIT-licenced Swift package that conforms to Apple's LanguageModel interface. Once integrated, developers can point an individual session at an SCX-hosted text model while continuing to use Apple's existing LanguageModelSession, @Generable, tool and streaming APIs.

The package supports text responses alongside light, moderate and deep reasoning levels. In practice, a single application can delegate one task to Apple's on-device model and hand another to an SCX-hosted model, sparing teams the work of engineering around a second, parallel AI framework. Because available models and their capabilities are resolved at runtime, new text models can be switched on without developers needing to update the package.

Project MAGPiE anchors the Australian angle. SCX describes it as a sovereign model served locally, offering a 131,000-token context window suited to tasks such as working with long documents. The runtime resolution mechanism means the Australian-hosted model can be surfaced to Apple developers as capabilities evolve, without any package-side changes.

SCX founder and CEO David Keane framed the move as a wager on Apple turning AI into baseline developer infrastructure. "Apple is making AI part of the basic plumbing of software development," he said, arguing that when millions of developers start building AI directly into the apps people use every day, "the amount of inference required is going to grow enormously." Keane added that most users will not necessarily know whether the AI they are using is running on their phone or hosted elsewhere — a blurring of on-device and cloud boundaries that the package's hybrid routing directly enables.

Tags: SCX.ai · SCXSwiftKit · Project MAGPiE · Apple · David Keane · Foundation Models


Leituras relacionadas


萬安算交所 AIXX · AIXX · Exchange verificado · Ir ao mercado