GPT-6 Astra (Low): benchmark results
Provider: OpenAI. Released 2026-09-03. Access: API.
Aggregate Rating 1918 ± 9, rank #34 of 2257 rated models, from 211 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| QuantArena - Iterative - Research Period | 58.1 | Score (%) | 100 |
| TavernBench | 63 | Play accuracy: share of tasks answered correctly, mean over | 100 |
| TavernBench - Rules | 95 | Accuracy on the ten rule-execution scenarios, eight tasks ea | 100 |
| ParseBench - Tables | 93.18 | GTRM Score (GriTS + TableRecordMatch) | 98.3 |
| JFinBench - No Tools - Financial Counter | 100 | Score (%) | 97.2 |
| AA GDP.pdf - Construction - Criterion Pass Rate | 78.97 | Criterion Pass Rate (%) | 96.9 |
| EnigmaForge - Intuition | 80.1 | Task success (%) on implicit-condition instances, where the | 96.9 |
| AA Omniscience | 40.55 | Score | 96.6 |
| AA GPQA Diamond | 93.13 | Accuracy (%) | 96.1 |
| AA GDP.pdf | 30.4 | All Criteria Pass Rate (%) | 96 |
| Korean CSAT 2026 (Easy Mode) - Physics I | 50 | Points (out of 50) | 96 |
| AA GDP.pdf - Real Estate - Criterion Pass Rate | 77.67 | Criterion Pass Rate (%) | 95.8 |
Interactive version: theaggregate.ai/model?slug=gpt-6-astra-low · How It Works · Data refreshed daily, snapshot 2026-11-17.