Google reports that Gemini 4 Carbon achieved 97% of the 656 exploitation capabilities in ExploitBench, using a best-of-three union in the autonudge setting.
Tasks: 41 V8 challenges with 16 capabilities each (656 in total); autonudge setting.
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. Best-of-three union is not a single-attempt score.
Google DeepMind
