HAL USACO — leaderboard

Princeton HAL leaderboard for USACO: 307 competitive programming problems from the USA Computing Olympiad across 4 difficulty tiers. Tests algorithmic problem-solving ability.

Metric: Accuracy (%). Source: hal.cs.princeton.edu. Status: saturation imminent. 13 models tracked.

Top models

#ModelScore
1GPT-5 (Medium)69.06
2O4 Mini (High)64.82
3O4 Mini (Low)53.09
4Claude Opus 4.1 (High)51.47
5Claude Opus 4.1 (20250805)48.21
6O3 (Medium)46.25
7GPT-4.1 (2025-04-14)44.95
8DeepSeek V3 (0324)39.09
9DeepSeek R138.11
10Claude 3.7 Sonnet (20250219)29.32
11Gemini 2.0 Flash (001)27.04
12Claude 3.7 Sonnet (High)26.71

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