A “digital twin” sounds like a second brain in a box. Meta’s TRIBE v2 is narrower: a model that predicts fMRI activity in response to media, giving researchers a way to test some hypotheses without recruiting people for every experiment.
What TRIBE v2 predicts
From media to fMRI: TRIBE v2 predicts how the brain responds to sights, sounds, and language. Meta says it can make zero-shot predictions for new subjects, languages, and tasks, and that it consistently outperforms standard modeling approaches.
The model was trained using data from more than 700 healthy volunteers who were shown a range of material, including images, podcasts, videos, and text. That marks a change from Meta’s earlier Algonauts 2025 award-winning model, which was trained on low-resolution fMRI recordings from four people.
Meta describes TRIBE v2 as delivering 70 times the resolution of similar models. The announcement does not name the comparison models or explain how that figure was measured, so the claim is useful context, not a complete account of performance.
A research tool, not a replacement brain
The practical pitch is that neuroscientists and clinical researchers can use predictions to test ideas about brain responses without running a human experiment for each one. That could make some early-stage research faster. It does not mean the model replaces participants, nor that it explains why the brain responds as it does.
“Digital twin” can imply a full replica. The mechanics described here are more specific: researchers provide media, and the model predicts fMRI activity. Those predictions may help guide experiments, but a prediction is not a direct measurement of a person’s brain or proof of an underlying mechanism.
Meta also points to possible uses beyond neuroscience, including drawing on brain research to inform AI systems. It frames computational simulation as a way to support research into neurological disorders, but the announcement describes a research opportunity, not a demonstrated treatment or clinical outcome.
Weights, code, and a demo
Open for research: Meta says it is sharing a research paper, model weights, and code under a CC BY-NC license. An interactive demo is also available for people to try TRIBE v2.
The release gives researchers a model to examine and test, rather than asking them to take a broad claim about a digital brain on faith. The meaningful question now is where its predictions help researchers ask better questions, and where a real scan remains the more useful answer.