SecCodeBench: Results as Published

SecCodeBench tests whether code from a model or a coding agent is both functional and secure. It has 98 runnable project tasks drawn from real vulnerabilities, in Java, C and C++, Python, Go, and Node.js, covering 22 CWE types, in both code generation and code repair. Code must pass the functional tests before any security check. Security is checked first by running an exploit, and otherwise by a judge model, and the published score is a weighted pass@1 that gives more weight to the setting with no security hint and to more severe weaknesses.

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.