Holo4

Last reviewed October 2026

Context

N/A

Pricing

Not publicly disclosed

Modalities

image, video, text

Released

Sep 2026

Overview: Holo4 is World Labs' fourth-generation spatial intelligence model, designed to generate and reason over interactive 3D environments from images or text prompts. It represents the core technology that drove AMD's $8.2 billion acquisition of World Labs in 2026, positioning spatial world generation as production-grade infrastructure rather than a research artifact.
Why it matters: World models that can simulate 3D environments are increasingly recognized as foundational to robotics, autonomous systems, and agentic AI that must reason about physical space — not just language. Holo4's capabilities directly address the gap between current LLMs and embodied AI systems that need to plan and act in the real world. AMD's acquisition signals that world model inference will be a major driver of next-generation compute demand, putting Holo4 at the center of the chip vendor's competitive response to Nvidia's physical-AI ecosystem. For enterprise buyers in manufacturing, robotics, and simulation, Holo4 represents the first commercially-backed world model with a major hardware sponsor's R&D weight behind it.

Key strengths

Generates interactive, explorable 3D worlds from single images or text descriptions
Spatial reasoning over scene geometry, depth, and object relationships
Designed for real-time interaction, not just static scene generation
Backed by AMD's hardware ecosystem, enabling optimized inference on AMD MI-series GPUs
Led by Fei-Fei Li's team, bringing ImageNet-era visual understanding expertise to world modeling

THE FRIDAY BRIEFING

We cover ai models every week.