CritPt: Results as Published

CritPt tests whether a language model can solve research-level physics problems of the kind a working physicist meets in a research project. The benchmark has 71 composite research challenges, each split into simpler checkpoint tasks (190 in total), across areas such as condensed matter, quantum physics, astrophysics, high energy physics and biophysics. All problems are new and were written by more than 50 active physics researchers from their own research. Each answer is hand-curated to be guess-resistant and machine-checkable, and the main metric is average accuracy, shown on Artificial Analysis as the 'CritPt: Score' percentage.

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.