Apple will headline the International Conference on Learning Representations in Rio de Janeiro, April 23‑27, with almost 60 research papers and hands‑on demos. Key attractions include the SHARP model demo, which rebuilds photorealistic 3‑D scenes from a single image in under a second, showcased on an iPad Pro powered by the new M5 chip.

At booth #204, Apple will also reveal an on‑device large‑language‑model demo running on a MacBook Pro with the M5 Max, leveraging its MLX framework. The quantized frontier‑coding model executes entirely locally inside Xcode, illustrating Apple’s push toward edge‑AI that keeps data on the device.

The conference schedule lists posters, oral sessions, and workshops that cover a broad swath of AI research, from generative models to reinforcement learning. Apple’s contributions will demonstrate how its silicon and software stack accelerate model inference, a trend that could reshape performance benchmarks across the industry.

By showcasing these technologies at ICLR 2026, Apple signals its intent to dominate the AI‑on‑device space and offers developers a concrete pathway to deploy sophisticated models without cloud dependence. The company’s emphasis on hardware‑software synergy sets a new standard for mobile AI performance.