GPT-6 (xHigh): benchmark results

Provider: OpenAI. Released 2026-09-03. Access: API.

Unified ELO 1811 ± 1, rank #3 of 1761 rated models, from 25 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA GPQA Diamond96.26Accuracy (%)100
DeepSWE74.1Pass@1 (%)100
DeepsecBench37.79Recall-weighted F2 score (%)100
Artificial Analysis Intelligence Index54.31Intelligence Index99.7
AA CritPt31.43Accuracy (%)99.6
AA Omniscience43.42Score99.6
AA Omniscience - Software Engineering (SWE)91Accuracy (%)99.6
ARC-AGI-293.33Accuracy (%)99.5
ARC-AGI-198.5Accuracy (%)99.3
AA MMMU-Pro86.24Accuracy (%)99.2
AA Humanity's Last Exam54.59Accuracy (%)99
AA Omniscience - Health53.7Accuracy (%)98.9

Interactive version: theaggregate.ai/model?slug=gpt-6-xhigh · How It Works · Data refreshed daily, snapshot 2026-09-05.