GPT-6 Astra (xHigh): benchmark results
Provider: OpenAI. Released 2026-09-03. Access: API.
Unified ELO 1988 ± 11, rank #8 of 2072 rated models, from 122 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA GDP.pdf | 32.2 | All Criteria Pass Rate (%) | 100 |
| AA GPQA Diamond | 96.26 | Accuracy (%) | 100 |
| Bug Hunt Bench - VS Code Extension | 23 | Planted Bugs Fixed (out of 45) | 100 |
| DeepSWE | 74.1 | Pass@1 (%) | 100 |
| DeepSWE-mini (16 Tasks) | 81.25 | Pass@1 (%, 16-task subset) | 100 |
| DuelLab Overall | 94.8 | DuelLab Score | 100 |
| Epoch AI - Weirdml V3 | 42.22 | Score | 100 |
| LLM2014 Logic 2026-09 | 88.75 | Median Score | 100 |
| LLM2014 Logic 2026-10 | 87.82 | Median Score | 100 |
| NonoBench | 100 | Overall Accuracy (%) | 100 |
| SEAL - SWE Atlas - Refactoring | 59.05 | Score | 100 |
| SciUniverse - Analyze XRD patterns | 82.5 | Pass@1 (%) | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-6-astra-xhigh · How It Works · Data refreshed daily, snapshot 2026-10-02.