CERN, the European particle physics laboratory, and the French AI company Mistral AI agreed on October 15, 2026 to build AI foundation models for science. Their first project is a Large Physics Model trained on billions of simulated particle collisions from the Large Hadron Collider. It matters because physics is about to produce far more data than today's tools were built for, and the partners want one general model to replace many narrow ones.
What was signed
The two sides signed a statement of intent for a long-term strategic partnership, CERN's announcement says. That is a plan to work together, not a contract with set funding. Reuters reports that neither side disclosed a timeline or a budget.
A foundation model is a large AI model trained once on a huge amount of data, then reused for many different tasks. CERN calls the Large Physics Model the first foundation model for collider physics.
What the Large Physics Model will do
The model will be trained on billions of simulated proton-proton collisions at the LHC, the world's largest particle accelerator. CERN lists four jobs for it:
- Reconstructing collisions, meaning working out what happened from the signals detectors record.
- Identifying particles that come out of each collision.
- Flagging unexpected signals that could hint at new physics.
- Evaluating designs for future detectors.
Today each of these jobs uses its own specialized algorithm. Reuters reports that the Large Physics Model is meant to replace them, and that the models could later serve other sciences.
The two partners split the work. CERN brings its physics expertise, scientific data and large-scale computing. Mistral brings its experience in training large models, CERN says.
Why physics needs it
The data problem is about to grow. CERN says its LHC experiments have stored around one exabyte of data so far, about a million terabytes. The High-Luminosity LHC, an upgrade expected to start around 2030, will raise the number of collisions tenfold.
"High-energy physics provides a unique environment in which artificial intelligence can be developed and tested," said Enrica Porcari, CERN's Chief Information Officer. Johannes Brandstetter, Vice-President of AI for Engineering at Mistral AI, said the company is "proud to work with CERN to explore what large-scale AI models can bring."
The deal builds on talks that go back to 2025, Crypto Briefing reports. It comes days after Mistral previewed Mistral Large 4, a model with about a trillion parameters. CERN has not said which Mistral model, if any, the physics work will start from.
What this means for developers
For machine learning engineers, the project is a test of whether one pretrained model can beat many task-specific ones in a field with strict accuracy needs. If it works, the same approach could reach other areas with huge sensor data, such as astronomy, climate or medical imaging.
For teams that build scientific software, watch how CERN shares the results. CERN has leaned on open software before: last month it moved 2,200 industrial computers to Debian. Neither partner has said whether the Large Physics Model's weights will be open, so that is the key question to ask.
For European buyers of AI, the deal shows Mistral courting public research bodies, not only companies.
There are no results yet, and no dates. Watch for the first technical paper, which will show how the model compares with the algorithms it is meant to replace.
This article was first published on Tech AI Wire.
Also available in
Deutsch · 日本語 · Français · Español · Português
Related on Tech AI Wire
- Mistral Large 4 launches with 1.05T parameters
- CERN migrates 2,200 industrial computers from RHEL to Debian 13
Sources
- CERN and Mistral join forces to accelerate AI for science - CERN
- CERN and Mistral partner to build AI foundation models for physics - Reuters via KSEV Radio
- CERN partners with Mistral to build a Large Physics Model - Crypto Briefing
Top comments (0)