Benchmark profile
xAI long-horizon jailbreak compliance (Long-horizon jailbreak compliance)
The rate at which a model complies after a multi-turn or long-horizon jailbreak attempt.
Data verified 26 confirmed releases in the last 30 daysStart free briefBenchmark score on Long-horizon jailbreak compliance — August 12, 2026
BenchLM mirrors the published score view for Long-horizon jailbreak compliance. Grok 4.6 leads the public snapshot at 1.0%. We do not use these results to rank models overall.
Benchmark score table (1 model)
ScoreAbout Long-horizon jailbreak compliance
Year
2026
Tasks
Long-horizon jailbreak conversations
Format
Compliance rate
Difficulty
Adversarial multi-turn safety
No matching public task set or result page was found. Lower compliance is better; the lane is display-only.
BenchLM freshness & provenance
Version
Long-horizon jailbreak compliance 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
FAQ
What does Long-horizon jailbreak compliance measure?
The rate at which a model complies after a multi-turn or long-horizon jailbreak attempt.
Which model scores highest on Long-horizon jailbreak compliance?
Grok 4.6 by xAI currently leads with a score of 1.0% on Long-horizon jailbreak compliance.
How many models are evaluated on Long-horizon jailbreak compliance?
1 AI models have been evaluated on Long-horizon jailbreak compliance on BenchLM.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 2,000+ readers.
One email each week. Unsubscribe anytime.