What open intelligence means
- 1
Model Weights
We will release open weights for our models.
- 2
Science
We will publish the research behind our models as papers, including technical reports that describe our methods.
- 3
Software
We will open source software that lets the community customize and build with our models, including reinforcement learning tools and environments.
Open intelligence powers progress
Some of the most important computing platforms in history emerged from open collaboration and shared discovery, from the internet and Linux to the foundational frameworks powering modern machine learning today. Open systems evolve faster, attract broader participation, and foster ecosystems of research and innovation that no single organization can achieve alone. They’re also safer: when systems are open to inspection, more researchers can identify vulnerabilities, improve safeguards, and contribute to their long-term resilience.
As AI rapidly becomes foundational for the modern world, infrastructure this important should be open, adaptable, and accessible to the people building with it.
By building in the open, we ensure that AI is not controlled by a few, but becomes the foundation upon which everyone can build the next era of progress.
Our convictions
Openness drives progress
Open systems drive faster innovation, better outcomes, and broader participation in the future of AI. When intelligence is shared, progress moves faster and benefits more people.
Safety demands scrutiny
When models are closed, safety research is bottlenecked by a few labs. When models are open, researchers can inspect weights, probe for risks, and make these systems safer.
Power to the builders
Researchers, developers, and organizations should be able to deploy, govern, and evolve AI systems on their own terms. AI is a powerful technology and, rather than concentrating it in the hands of a few, it should be diffused.