A new machine may help physicists decide which signals deserve a second look. On October 15, 2026, CERN announced a planned partnership with Mistral to develop foundation models for science.
Learning the language of collisions
According to CERN’s announcement, the proposed Large Physics Model would train on billions of simulated proton collisions. Its intended jobs include reconstructing collisions, identifying particles, spotting unexpected signals and informing detector designs.
Current reconstruction uses successive specialised algorithms. The ambition is a model that can work across multiple stages. CERN also reports storing roughly one exabyte of LHC experimental data.
An announcement, with experiments still ahead
This is a statement of intent, not a demonstrated breakthrough or a discovery of a new particle. Apep Bar’s assessment: the interesting test will be whether broader AI models can produce results researchers can independently check.
If a machine flags something unexpected, what evidence will persuade physicists that nature—not the model—is responsible?
Featured image: abstract illustration by Dynamic Wang / Unsplash; not a CERN detector image.
Question. Investigate. Reconsider.
