LiblibAI

Trains industrial robots using AI models trained on public videos, enabling robots to understand object motion and real-world behavior

Series A · $450M · 2026-03 · Valuation $1.7B

Trains "large action" foundation models and general-purpose agents with spatiotemporal reasoning capabilities on action-labeled gameplay footage, and transfers them to real-world embodied systems such as robots, drones and autonomous vehicles.

Undisclosed · $650M · 2026-09 · Valuation $6.2B

Multimodal medical AI models trained on diverse clinical data

Seed · $9M · 2025-09

Develops multimodal image/video generation large models (HiDream series) and is evolving toward native omni-modal world models

Series C · $208M · 2026-07

Develops 3D-native foundation models (Tripo series) and 3D data infrastructure, and provides tools and a platform for generating 3D models from text/images (Tripo Studio), serving creators and developers worldwide in gaming, animation, robotics, industrial design and other use cases.

Series B · $694M · 2026-09 · Valuation $1.0B

Develops multimodal large models and world models, and has launched products including Vidu

Series B · $586M · 2026-04

Trains foundation models that natively generate precise, controllable vector graphics, producing editable charts, infographics, diagrams, maps and brand layouts from text, images or existing graphics

Pre-Seed · $5M · 2026-09

Develops on-device embodied brains, embodied multimodal large models, spatial intelligence, and embodied interactive world models for the real physical world

$139M · 2026-06

Multimodal foundation world models and physics simulation environments, providing scalable interactive training grounds for robots and other embodied agents

Seed · 2026-08