O3 Pro (High) — benchmark results

O3 Pro evaluated at the high reasoning-effort setting. Provider: OpenAI. Released 2025-06-10. Access: API.

Unified ELO 1914 ± 95, rank #56 of 1776 rated models, from 8 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Ducky Bench (Stabby Quack)1436ELO100
Aider polyglot coding leaderboard84.9Pass rate (%)97
Visual-Language Understanding51.63Score (self-reported)90.6
SEAL - VISTA51.63Score90.3
WeirdML58.21Average Score75.9
SEAL - MASK82.5Score68.2
ARC-AGI-159.33Accuracy (%)58.4
ARC-AGI-24.86Accuracy (%)47.5

Interactive version: theaggregate.ai/model?slug=o3-pro-high · How the rankings work · Data refreshed daily, snapshot 2026-07-22.