
Apple launches foundation models for developers, with 7B text and 2B code models
Apple unveiled two foundation models—a 7‑billion‑parameter text generator (AppleGPT‑3) and a 2‑billion‑parameter code model (AppleCode‑2)—through a new REST API, with on‑device inference support and pricing that undercuts major cloud providers.
Apple opened access to its first suite of foundation models, a 7‑billion‑parameter text generator (AppleGPT‑3) and a 2‑billion‑parameter code model (AppleCode‑2), via a new REST API [Apple Newsroom]. The API endpoint https://api.apple.com/foundation/v1/ offers POST /generate for text completion and POST /code for code synthesis. Requests take JSON with model, prompt, and optional temperature; responses stream token payloads. Pricing is tiered: the first 1 M tokens per month are free, then $0.0005 per token for text and $0.001 per token for code. Apple bundles the service with Xcode 15.2 and provides a Swift SDK that routes inference to on‑device Apple silicon when the model fits in GPU memory, otherwise falling back to Apple‑hosted cloud nodes. Documentation lists latency of 120 ms for 256‑token completions on an M2 Max, versus 210 ms on comparable AWS instances [hn-front]. The service integrates with Apple’s privacy framework, ensuring prompt data remains on the device unless the developer opts into cloud processing.
Why it matters
Apple now competes directly with OpenAI, Anthropic, and Google in the foundation‑model market, becoming the first major consumer‑hardware vendor to ship its own LLM platform. Tight integration with Apple silicon enables the 7B model to run entirely on‑device on M2 Ultra, cutting latency and eliminating reliance on network bandwidth for offline AI features. The pricing undercuts $0.002‑per‑token rates of many public APIs, reshaping cost calculations for mobile‑first products.
Editor's take
Apple’s entry is less about democratizing AI than about cementing a proprietary stack. By tying model access to its hardware and developer tools, Apple creates a high‑performance but siloed ecosystem that forces iOS/macOS teams to choose between cross‑platform flexibility and the latency‑price advantage of native Apple models. Competitors will have to match not only raw capability but also the seamless hardware‑software coupling Apple now offers.
Subscribe to the broadcast.
Daily digest of the day's most important tech news. No fluff. Engineering signal only.
// delivered via substack · double-opt-in confirmation


