GPT-6 Astra (High): benchmark results
Provider: OpenAI. Released 2026-09-03. Access: API.
Unified ELO 1976 ± 11, rank #11 of 2072 rated models, from 118 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA-Omniscience Index - Health | 27 | Omniscience Index | 100 |
| AI for Education Pedagogy | 92.66 | Accuracy (%) | 100 |
| BVB | 70.07 | Overall (%) | 100 |
| Computer Anthology Terminal Tasks (Codex CLI) | 65.6 | pass@1 (%) | 100 |
| FWBench | 0.4 | Decision score S (lower is better; 0 is the hindsight-optima | 100 |
| HLE-Diamond (With Tools) | 82.9 | Accuracy (%, web + code tools) | 100 |
| Korean CSAT 2026 (Easy Mode) - Total | 450 | Points (out of 450) | 100 |
| LLM Chess (Saplin) | 1613.8 | ELO | 100 |
| OpenAI GPT-6 Sol & Luna Launch - Factual Errors on User-Flagged Conversations | 3.9 | Answers with any factual error (%) | 100 |
| Taiwan Exams - GSAT (ROC 115) | 98.81 | Points Scored (%) | 100 |
| AA-Omniscience Index - Software Engineering (SWE) - JavaScript | 88.18 | Omniscience Index | 99.9 |
| AA Omniscience | 43.73 | Score | 99.8 |
Interactive version: theaggregate.ai/model?slug=gpt-6-astra-high · How It Works · Data refreshed daily, snapshot 2026-10-02.