Apple showcases 60 AI studies at ICLR 2026 Apfelpatient

Apple is sponsoring the International Conference on Learning Representations, taking place from April 23-27 in Rio de Janeiro, where it will present nearly 60 research projects and technology demonstrations. Among the highlights is the SHARP model, which creates 3D scenes from individual photos in under a second.

Apple is intensifying its public presence in AI research. Just recently, Apple announced it would be presenting several new research papers at the CHI conference. Now comes its next major appearance: at the International Conference on Learning Representations (ICLR) 2026, Apple will present nearly 60 research papers and a range of technical demos.

The ICLR is one of the most important conferences for machine learning and takes place from April 23 to 27 in Rio de Janeiro. Apple is not only a sponsor but will also be showcasing several live demonstrations of its latest AI models at its booth (number 204) during exhibition hours. This underscores Apple's strategy to increase its visibility within the research community – despite, or perhaps even because of, the recent organizational changes in its AI division.

SHARP: 3D scenes made from individual photos

One of the most impressive demo highlights is the SHARP model. It can reconstruct a photorealistic 3D scene from a single 2D photo in under a second – a so-called Gaussian splatting rendering that can be viewed from any angle.

Apple is running the demo live on an iPad Pro with an M5 chip. The fact that such a computationally intensive model works on a tablet demonstrates the power of the current Apple Silicon generation – and provides a glimpse into the kind of AI applications Apple intends to enable directly on its devices in the future, without relying on cloud processing.

On-Device LLM on the MacBook Pro

A second demo shows on-device LLM inference on a MacBook Pro with an M5 Max using MLX. MLX is Apple's open-source framework specifically designed for AI inference on Apple Silicon.

The demo runs a quantized frontier coding model entirely locally in Xcode – the development tool provided by Apple itself. This is an interesting alternative to cloud-based coding assistants like Claude Code and OpenAI Codex : Apple demonstrates that powerful AI support can also function completely locally.

What that signals

Apple is often accused of lagging behind OpenAI, Google, and Anthropic in AI development. Its presence at ICLR reveals a different side: Apple is conducting intensive research, regularly publishing papers, and developing technologies that focus particularly on integration with end devices. This distinguishes Apple's approach from that of the major cloud AI providers.

The demos at ICLR and the CHI presentations on AirPods Pro and Apple Intelligence indicate where Apple sees its AI strengths: on-device models, spatial computing, and multimodal applications seamlessly integrated into Apple devices. Whether these research findings can be translated into concrete products remains an open question—especially for the revamped Siri, expected at WWDC in June. (Image: Shutterstock / Pressmaster)