FullStackBench: Results as Published

FullStack Bench tests whether a model can write code for full-stack programming work, not only short algorithm problems. It has 3,374 problems in 16 programming languages, and each problem exists in both Chinese and English (1,687 each), across domains such as basic programming, data analysis, software engineering, mathematics, and machine learning. Each item includes a question, unit tests, a reference solution, and labels. A sandbox runs the tests, and the reported score is the pass rate, including test Pass@1.

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.