LLM Stats (CyberGym): leaderboard

Metric: Score (%). Source: llm-stats.com. 15 models tracked.

Top models

#ModelScore
1GLM-5.384.5
2DeepSeek V4 Pro (0813)83.3
3Claude Mythos Preview83.1
4GPT-5.581.8
5Claude Opus 4.878.8
6Hy4 preview78.4
7DeepSeek V4 Flash (0731)76.7
8Claude Opus 4.673.8
9Claude Opus 4.773.1
10Seed 2.1 Pro70.2
11GLM-5.168.7
12Seed 2.1 Turbo67
13Kimi K2.541.3

Interactive version: theaggregate.ai/benchmark?slug=llm-stats-cybergym · How It Works · Data refreshed daily, snapshot 2026-09-05.