Models & LLMs

Odyssey-3: A Free Generative World Model for Exploration and Control

Odyssey-3 is a new generative world model that allows users to create interactive environments from text prompts. It can control robots, drones, and video games, with public access and API availability.

The Decoder Β· Oct 11, 2026

What happened

  • Odyssey-3 generates interactive environments from text prompts in real time.
  • Odyssey-3 can control robots, drones, and video game characters.
  • Odyssey-3 has top scores on physics benchmarks but with caveats.

Why it matters

Odyssey-3 represents a significant advancement in generative world models, offering tools for researchers and developers to explore interactive environments. Its potential applications in robotics, autonomous driving, and video games highlight its versatility and impact on AI research.

The Elephant take

🐘 ιΌ‹ Odyssey-3 is a fascinating glimpse into the future of AI, but the physics benchmark results are a reminder that even the most advanced models have limitations. The free demo is a great starting point for exploring this new frontier.

Who should care

  • AI researchers
  • Robotics developers
  • Video game creators

What to do next

  1. Explore the free online demo
  2. Apply for API access
  3. Review the benchmark results carefully

Keep in mind

The physics benchmark scores come with caveats and may not reflect the model's true capabilities.

Read the original reporting at The Decoder β†—