GPT-6 Astra (Medium): benchmark results
Provider: OpenAI. Released 2026-09-03. Access: API.
Aggregate Rating 1927 ± 8, rank #28 of 2257 rated models, from 261 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AI Coding Daily (Codex CLI) - Python Code Quality | 18.5 | Python Code Quality (max 20) points, LLM-judged rubric score | 100 |
| Computer Anthology Terminal Tasks (Terminus-2) | 67 | pass@1 (%) | 100 |
| DistDebug-Bench - Block | 67.7 | Coverage (%) | 100 |
| DistDebug-Bench - Line | 53.9 | Coverage (%) | 100 |
| DistDebug-Bench - Mention | 82.8 | Coverage (%) | 100 |
| DrivingBench | 100 | Course progress (%, best of up to 3 attempts in one chat) | 100 |
| DrivingBench (First Attempt) | 48.82 | Course progress (%, first attempt) | 100 |
| Maze-Bench | 94.99 | Mean Score | 100 |
| OpenVLHarness Study - Base - Average | 60.27 | Published Task Score (%) | 100 |
| OpenVLHarness Study - Base - BC-VL | 46.12 | Published Task Score (%) | 100 |
| OpenVLHarness Study - Base - CountQA | 82 | Published Task Score (%) | 100 |
| OpenVLHarness Study - Base - ERQA | 76.5 | Published Task Score (%) | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-6-astra-medium · How It Works · Data refreshed daily, snapshot 2026-11-17.