PixVerse R1 World Model Powers Game Engine Following $439M Round
PixVerse secured a $439 million Series C extension to scale its real-time generative game engine, pushing the video AI startup's valuation over $2 billion.
Singapore-based AI startup PixVerse has closed a $439 million Series C extension, pushing its valuation past $2 billion. The funding announcement coincides with the launch of the PixVerse Game Engine, a platform that shifts the company from static video generation to real-time interactive environments. This capitalizes on the momentum of the R1 world model the company released in January.
The Series C extension brings in new investors including Alibaba, Mirae Asset, and BlueFocus. It follows an initial tranche led by CDH Investments in March 2026 that originally valued the parent company, AIsphere, at approximately $1 billion.
The PixVerse Game Engine
The new game engine architecture relies entirely on the R1 world model rather than traditional rendering pipelines. The engine separates abstract game mechanics from visual expression. Players use natural language to interact with the environment, and the system generates both the mechanical logic and the visual response simultaneously.
CEO Wang Changhu designed the engine to generate worlds continuously based on user actions, moving away from pre-rendered assets. This approach requires heavy compute infrastructure to maintain low latency during real-time generation, creating a different set of challenges than traditional AI inference tasks.
PixVerse attributes its scaling success to a proprietary data labeling methodology. Leadership highlighted techniques adapted from Wang’s previous work optimizing recommendation algorithms at TikTok, focusing on data quality over raw dataset scale to train the core models.
Model Tiers and Performance
PixVerse segments its models into three distinct product tiers targeting different developer and enterprise workloads. The flagship V6 model currently supports up to 4K resolution with integrated audio generation.
| Series | Target Workloads | Core Capability |
|---|---|---|
| V-Series | Consumer and API | 4K resolution with integrated audio (V6) |
| C-Series | Professional workflows | Commercial film integration |
| R-Series | Interactive entertainment | Real-time world modeling (R1) |
The platform currently supports 150 million registered users and processes 15 million monthly active users across 177 countries.
PixVerse prices its image-to-video API at $4.80 per minute of generation. Developers integrating the API into consumer applications will need to carefully monitor user behavior to reduce LLM API costs as usage scales. As the company pushes deeper into interactive entertainment, the transition from discrete video generation requests to continuous interactive streams will fundamentally change how developers calculate infrastructure spend.
Get Insanely Good at AI
The book for developers who want to understand how AI actually works. LLMs, prompt engineering, RAG, AI agents, and production systems.
Keep Reading
How to Fine-Tune Cosmos Predict 2.5 for Robotics With LoRA
Learn how to adapt NVIDIA's 2B and 14B Cosmos Predict 2.5 world foundation models using parameter-efficient fine-tuning methods like LoRA and DoRA.
Runway Announces $10M Fund for Early-Stage AI Startups
Runway formalizes its venture arm with a $10 million fund and Builders program to support early-stage startups using its video intelligence infrastructure.
Varya 14B Distills Wan 2.2 for $0.005/Sec Video Generation
Avataar AI has launched Varya, a 14-billion-parameter open-weight video model distilled from Wan 2.2 that cuts generation costs to $0.005 per second.
Decart Oasis 3 API Renders Endless Driving Sims at 22 FPS
Decart's Oasis 3 is an interactive world model available via API that generates real-time, closed-loop driving environments for autonomous vehicle validation.
Origin Lab Raises $8M for Game Engine Telemetry Marketplace
Origin Lab has secured $8 million in seed funding to launch a platform that converts raw video game engine data into licensed datasets for world model research.