GPT-6: benchmark results

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

Unified ELO 1805 ± 1, rank #4 of 1392 rated models, from 30 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Benchmarks.bio - TxBench-AD56.9Pass Rate (%)100
LLM Stats (ARC-AGI-3)99.9Score (%)100
LLM Stats (Agents' Last Exam)59.3Score (%)100
LLM Stats (BrowseComp)91.5Score (%)100
LLM Stats (ExploitBench)100Score (%)100
LLM Stats (ExploitGym)42.4Score (%)100
LLM Stats (ScreenSpot Pro)92.7Score (%)100
LLM Stats (Terminal-Bench 4.0)57.7Score (%)100
LLM Stats Score60.72LLM Stats Score (conservative rating)100
RuneBench34063Total Peak XP Rate (XP/min)100
Terminal-Bench 4.058.2Accuracy (%)100
Vellum - ARC-AGI-295Score (%)100

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