Anthropic AI Solved Nine-Loop Physics Calculation
Physicists used the Claude model to surpass the prior eight-loop scattering amplitude record set in 2023.
Updated on Sept. 25, 2026 in Physics
Live Poll
Do you believe the use of AI for complex scientific research will improve future discovery?
Physicists Liam Fitzpatrick and Siddharth Mishra-Sharma utilized the Claude AI platform to successfully compute the nine-loop six-particle scattering amplitude. The research team released the resulting data files publicly on Zenodo following the calculation.
Why it matters
This development demonstrates the capacity for large language models to perform complex symbolic calculations in theoretical physics, moving beyond mere text generation. It accelerates research in planar N=4 super Yang-Mills theory by significantly reducing the human-led effort required for high-loop operations.
The team reached the nine-loop milestone for six-particle scattering in planar N=4 super Yang-Mills theory, generating over 100 megabytes of data files. The project required a total compute cost of only a few thousand dollars.
The players
Anthropic
An AI research company focused on building steerable and reliable large language models, including the Claude series.
Liam Fitzpatrick
A physicist specializing in high-energy theoretical physics and numerical methods.
Siddharth Mishra-Sharma
A physicist who contributed to the development of the nine-loop computation techniques.
Lance Dixon
A physicist known for his research on scattering amplitudes in quantum field theory.
Matt von Hippel
A researcher who initiated the public challenge for the nine-loop calculation.
The details
The computation employed two specific methods: a direct bootstrap of hexagon functions—a mathematical technique for constructing scattering amplitudes—and an indirect route through the nine-loop form factor. Claude operated largely unsupervised for several days, handling the symbolic manipulation required for these complex equations. Physicist Lance Dixon performed an independent verification of the results to ensure accuracy.
Timeline
August 7, 2026: Matt von Hippel issued the open AI challenge.
Late August or early September 2026: The nine-loop result calculation was completed.
September 16, 2026: The data files were released publicly on Zenodo.
September 25, 2026: Anthropic announced the nine-loop calculation results.
The Tech Race
This achievement marks a significant advancement in the competitive field of automated scientific discovery within quantum field theory. It extends the prior research program by pushing beyond the eight-loop limit established in 2023.
The availability of the 100-megabyte data set on Zenodo allows other theoretical physicists to integrate these high-loop findings into their own models immediately. This reduces the barrier to entry for researchers needing high-precision scattering data for quantum field theory simulations.
The takeaway
This breakthrough proves that large language models can act as reliable assistants for high-level symbolic mathematics in theoretical science. Researchers should watch for upcoming peer-reviewed papers that utilize this nine-loop dataset to refine standard model predictions.
Further reading
Explore deeper developments in the field within our dedicated Physics coverage.
Live Poll
Do you believe the use of AI for complex scientific research will improve future discovery?
