Skip to content
Spellwright News

Benchmarks

ExploitGym (default 6-hour timeout), pass@1

ExploitGym (default 6-hour timeout), pass@1

Other · higher is better · 1 model

Pass@1: the share of problems solved on the first try.

  1. Gemini 4 Carbon53.5%*

Every result

  • Gemini 4 Carbon53.5%Reported by the model's maker (Google), Nov 10, 2026

    Google reports 53.5% for Gemini 4 Carbon on ExploitGym at the default 6-hour timeout, in a chart in its safety report.

    Tasks: 869 tasks; default 6-hour timeout per task.

    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. The value was read from a labeled bar chart in the safety report.

    Google DeepMind · Google DeepMind

No published ExploitGym (default 6-hour timeout), pass@1 score checked yet for: Claude Fable 5.2, Claude Mythos 5.2, Gemini 4 Argon, Claude Sonnet 5.5, Claude Opus 5.5, Claude Opus 5, Claude Sonnet 5, Gemini 2.5 Pro Experimental (March 2025), Gemini 2.0 Flash (December 2024 experimental), Qwen2.5-7B, Qwen2.5-7B-Instruct, GPT-4o mini (July 2024 release).