Anthropic released a Swift package on June 9 for Apple’s Foundation Models framework, enabling developers targeting iOS 27, iPadOS 27, macOS 27, visionOS 27, and watchOS 27 to route complex workloads from Apple’s on-device model to Claude with no changes to session logic (Anthropic Blog). The package receives the typed Swift values produced by Apple’s @Generable annotations, delivering clean structured inputs to Claude rather than raw user text, and returns streaming responses, tool call results, and structured outputs directly to SwiftUI views (Anthropic Blog). Developers can swap the underlying provider from Apple’s on-device model to Claude, or to Google Gemini, by updating a single Swift Package Manager dependency with no modifications to session code (TechTimes). The release follows Apple’s WWDC 2026 keynote on June 8, which introduced the Foundation Models framework alongside the AI Extensions API for iOS 27 and Siri AI powered by a licensed Gemini model (Apple Newsroom).
OpenAI’s Dreaming V3 memory architecture launched June 4 for ChatGPT Plus and Pro subscribers in the U.S. and is rolling out to Free and Go tiers and additional countries over the coming weeks (9to5Mac); a roughly 5x reduction in per-request compute enabled the free-tier expansion (Android Headlines). The architecture keeps stored context current over time, automatically revising past-tense facts as circumstances change, and exposes a readable memory summary with per-topic user controls (TechTimes). Also on June 9, Google’s enterprise platform release notes record that Gemini 3.5 Flash became the non-optional default in the Gemini Enterprise app, with its management toggle removed across Global, U.S., and EU multi-regions (Google Cloud Docs); the model is priced at $1.50 per million input tokens and $9 per million output tokens with a 1 million-token context window (Google Blog).