CNB 3
A new foundation for protein design, discovery, and the next generation of biological research.
Biology is a search through an enormous space.
Proteins sit at the center of an enormous range of biological processes. Their sequences encode structure, interactions and function, but the space of possible proteins is far larger than any laboratory can explore directly.
Machine learning offers a way to navigate that space differently: learn representations from biological data, generate candidates, test hypotheses computationally, and return the most useful ideas to researchers. CaelisNeuralBase 3 (CNB 3) is built as infrastructure for that kind of work.
Introducing CaelisNeuralBase 3.
Inserloft Research has released CNB 3 as a public protein foundation model. It is designed to give researchers a starting point they can inspect, adapt, evaluate and build into specialized systems for protein design and biological discovery.
The model contains approximately 985.4 million parameters, uses the CNB3 architecture, and is released under the MIT license.
From a model to a research foundation.
The important step is not simply making a larger model. It is making a foundation that can become part of many experiments. CNB 3 can serve as the base layer for researchers developing their own protein-focused models, evaluations and computational workflows.
That makes the release intentionally open-ended. A foundation model can be specialized for a particular family, task or research question, while the underlying public checkpoint remains a shared starting point.
A new layer for Project Proteus.
CNB 3 extends the technical foundation of Project Proteus, Inserloft Research’s effort to use machine intelligence in the exploration of biological possibility.
Proteus is not a single model. It is a research loop: represent biological information, generate possibilities, evaluate them, learn from the results, and iterate. CNB 3 provides a new foundation inside that loop.
The release is the beginning, not the conclusion.
A foundation model becomes more valuable as researchers challenge it with new tasks, build specialized models from it, compare approaches and publish what they learn. CNB 3 is released with that future in mind.
We are building toward a world where biological research can explore more possibilities, move more quickly from hypothesis to experiment, and create new tools for understanding and improving life.
FOR A BETTER WORLD.