Skip to content

LLM Benchmark Hub

Model evidence, daily

Models What's New Trends Benchmarks Skill Maps
How It Works Metrics
LLM Benchmark Hub
Loading data...

Frontier Security Agents - Cybench Hard — leaderboard

Metric: Success rate (%). Source: arxiv.org. 8 models tracked.

Top models

#ModelScore
1GPT-5.594.1
2DeepSeek V4 Flash86.4
3GPT-5.6 Luna79.5
4Claude Opus 4.876.2
5GPT-5.6 Terra65.8
6DeepSeek V4 Pro43.9
7GPT-5.6 Sol9.4
8Claude Fable 50

Interactive version: aibenchmarks.dev/benchmark?slug=frontier-security-agents-cybench-hard · How the rankings work · Data refreshed daily, snapshot 2026-07-18.

Built by Mikhail Doroshenko — AI researcher, co-author of Humanity’s Last Exam. Independent project; no affiliation with any AI lab.