AISI Cyber TLO 10M: leaderboard

AISI cyber range: "The Last Ones", a 32-step corporate network attack simulation requiring reconnaissance, lateral movement, credential theft, reverse engineering, and data exfiltration. Avg steps completed at 10M token budget.

Metric: Avg Steps (/32). Source: arxiv.org. Saturation forecast: Around 2029. 10 models tracked.

Top models

#ModelScoreOverall rank
1GPT-5.510#31
2Claude Opus 4.69.8#63
3Claude Mythos Preview9.5#4
4GPT-5.48.8#79
5GPT-5.1 Codex8#159
6Claude Opus 4.57.6#83
7GPT-5.3 Codex7.2#69
8Claude Sonnet 4.56.1#142
9Claude 3.7 Sonnet5.8#249
10GPT-4o1.7#343

No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.

Interactive version: theaggregate.ai/benchmark?slug=aisi-cyber-tlo-10m · How It Works · Data refreshed daily, snapshot 2026-10-20.