Benchmark profile
CVE-Bench
A public benchmark for agents working with real-world software vulnerabilities.
Data verified 26 confirmed releases in the last 30 daysStart free briefBenchmark score on CVE-Bench — August 12, 2026
BenchLM mirrors the published score view for CVE-Bench. Grok 4.6 leads the public snapshot at 39.8% , followed by Grok 4.5 (35.2%). We do not use these results to rank models overall.
Grok 4.6
xAI
grok-4-6
Grok 4.5
xAI
grok-4-5
About CVE-Bench
Year
2026
Tasks
Real-world vulnerability tasks
Format
Pass@1
Difficulty
Advanced cybersecurity
A public leaderboard and GitHub implementation are available. BenchLM keeps the xAI capability-test rows display-only because the card reports a specific agent setup.
BenchLM freshness & provenance
Version
CVE-Bench 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 CVE-Bench measure?
A public benchmark for agents working with real-world software vulnerabilities.
Which model scores highest on CVE-Bench?
Grok 4.6 by xAI currently leads with a score of 39.8% on CVE-Bench.
How many models are evaluated on CVE-Bench?
2 AI models have been evaluated on CVE-Bench on BenchLM.
Compare Top Models on CVE-Bench
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.