Gemma 4 26B-A4B inference in ~2 GB of RAM on any M-series MacBook
Swift apple-silicongemmagemma4gemma4-26b-a4bgpgpullmllm-inferencelocal-aimacosmetalon-device-aion-device-llmswift
import Foundation
public protocol AppModelInstallerClient: Sendable {
var descriptor: AppModelInstallDescriptor { get }
func checkInstallRequirement(outputDirectory: URL) throws -> AppModelInstallRequirement
func installDefaultModel(outputDirectory: URL) -> AsyncThrowingStream<AppModelInstallEvent, Error>
func discardPartialInstall(outputDirectory: URL) async throws
func cancel()
}