Frontier Log

Benchmark · Active

ExploitGym

Created by Berkeley RDI, 11 May 2026 · Higher is better · Website

Settings

Task setScoringMitigationsReported by
ExploitGym 1.0 (869 tasks)ScoreMitigations offOpenAIGoogle
ExploitGym 1.0 (869 tasks)ScoreMitigations onOpenAI
ExploitGym 1.0 (869 tasks)ScoreNot specifiedOpenAIMetaDeepSeekMoonshot AI
ExploitGym 1.0 (869 tasks)Successful exploitsMitigations offAnthropicMetaZ.ai
ExploitGym 1.0 (869 tasks)Successful exploitsNot specifiedMeta

Bold: Benchmark default based on Paper and Leaderboard · Download image

Results

01020304050Oct 202636.94%GPT-6-Cyber29.69%GPT-6 Astra01020304050Oct 202636.94%GPT-6-Cyber29.69%GPT-6 Astra
ModelHarness and effortScoreSource
GPT-6-Cyber5 Nov 2026Harness not specifiedEffort not specified36.94%GPT-6 Cyber System CardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.hardened variant: ASLR and stack canaries on userspace targets, V8 sandbox on, stricter kernel configurations where supported
GPT-6 Astra3 Sep 2026Harness not specifiedEffort not specified29.69%GPT-6 Cyber System CardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.hardened variant: ASLR and stack canaries on userspace targets, V8 sandbox on, stricter kernel configurations where supported · reported later

All benchmarks