Google reports 88.2% pass@1 for Gemini 4 Carbon on the May 2026 SEC-bench Pro set of 183 JavaScript-engine vulnerabilities, averaged over three attempts.
Tasks: 183 historical JavaScript-engine vulnerabilities in Chromium (V8) and Firefox (SpiderMonkey).
Setting: Google says its results use the Gemini API at the highest thinking setting unless noted; no numeric budget is stated.
Limits: Provider-reported: Google publishes the result for its own model; it has not been independently reproduced here. Google says the scores are as of October 2026 for the model it calls Gemini 4 Carbon; equivalence to the build now served on every surface is not established. Results under the same benchmark name can be shown together as reported scores, but unmatched settings and missing methods rule out a controlled ranking. This record adds the benchmark version and settings Google gave on November 10; the October 28 record stays as history.
Google DeepMind · Google DeepMind · Google DeepMind · Google DeepMind
