GPT-6 Astra (Max): benchmark results

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

Aggregate Rating 1989 ± 5, rank #7 of 2257 rated models, from 868 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
4DCodeBench0.79Reconstruction score (0–1), five-family mean over 200 scenes100
AgenticBBO Frontier - BBOB f15 d100.66Normalized anytime/final score (0-1)100
AgenticBBO Frontier - Digits MLP HPO0.62Normalized anytime/final score (0-1)100
AgenticBBO Frontier - Sysbench Database Tuning0.5Normalized anytime/final score (0-1)100
AnalogLayoutWorld - DRC Pass Rate99.2Pass rate (%)100
AnalogLayoutWorld - Eligible Engineering Quality0.57Native mean score (0-1)100
AnalogLayoutWorld - Full Acceptance70.8Pass rate (%)100
AnalogLayoutWorld - LVS Pass Rate98.3Pass rate (%)100
AnalogLayoutWorld - Overall0.75Native mean score (0-1)100
AnalogLayoutWorld - Post-Layout Simulation Pass Rate82.5Pass rate (%)100
AutoRLBench - Issue45.5Root cause found, 33 failed training runs (%)100
AutoRLBench 32 Issue / 18 Solution46Pass rate over 32 issue judgments and 18 solution runs (%)100

Interactive version: theaggregate.ai/model?slug=gpt-6-astra-max · How It Works · Data refreshed daily, snapshot 2026-11-17.