Connecting experiments to computation
Recursion’s platform uses biological experiments and cell images to build data for AI-guided drug discovery. Its pipeline includes investigational candidates at different development stages. A model’s ability to prioritize experiments and a medicine’s clinical performance are separate parts of that process. [1]
A concrete model contribution
The Boltz-2 release credits researchers at MIT alongside Recursion. The model jointly predicts molecular structures and binding affinity, extending the kinds of questions a co-folding model can address. The release describes computational benchmarks, rather than a completed clinical validation. [2]
A public artifact that others can study
Boltz-2’s repository provides code and weights under the MIT license. This permits independent deployment and adaptation, distinct from buying access to Recursion’s broader discovery platform or datasets. Its research-team attribution should also be kept separate from assumptions about a later company bearing the Boltz name. [3][2]
Sources
- Pioneering AI Drug Discovery | Recursion
Recursion · retrieved Oct 23, 2026
- Introducing Boltz-2: Toward Accurate and Efficient Binding Affinity Prediction
Boltz team · Jun 6, 2025
- jwohlwend/boltz
Boltz team · retrieved Oct 23, 2026
- MIT and Recursion Release Boltz-2: Next Generation AI Model to Predict Binding Affinity at Unprecedented Speed, Scale, and Accuracy
Recursion Pharmaceuticals · Jun 6, 2025