HLCE — leaderboard

Humanity's Last Code Exam — 235 of the most challenging competitive programming problems from ICPC World Finals and IOI (2010-2024). Measures Pass@1 averaged across both contest types.

Metric: Pass@1 (%). Source: github.com. Status: saturation imminent. 12 models tracked.

Top models

#ModelScore
1O4 Mini (High)15.84
2Gemini 2.5 Pro (03-25)11.4
3O3 Mini (High)10.35
4O1 Mini5.95
5DeepSeek R14.83
6DeepSeek V3 (0324)3.53
7Claude 3.7 Sonnet2.48
8Claude 3.7 Sonnet (Thinking)2.12
9Claude 3.5 Sonnet1.71
10GPT-4o (2024-05-13)1.22
11GPT-4o ChatGPT1.18
12GPT-4o Mini0.48

Interactive version: theaggregate.ai/benchmark?slug=hlce · How the rankings work · Data refreshed daily, snapshot 2026-07-22.