GPT-6 Sol (Non-reasoning): benchmark results
Provider: OpenAI. Released 2026-09-22. Access: API.
Aggregate Rating 1767 ± 17, rank #215 of 2201 rated models, from 61 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MLCR-AA - Accuracy | 85.56 | Accuracy (%) | 88.3 |
| AA GDP.pdf - Construction - Criterion Pass Rate | 74.91 | Criterion Pass Rate (%) | 87.7 |
| AA GDP.pdf - Real Estate - Criterion Pass Rate | 73.33 | Criterion Pass Rate (%) | 87 |
| AA-Omniscience Accuracy | 45.18 | Accuracy (%) | 83.2 |
| Wolfram LLM Benchmarking Project | 59.9 | Correct Functionality (%) | 82.1 |
| ParseBench - Tables | 89.51 | GTRM Score (GriTS + TableRecordMatch) | 80.8 |
| MLCR-AA | 22.78 | Overall Score (%) | 79.2 |
| AA Omniscience | -0.85 | Score | 73.1 |
| ParseBench - Semantic Formatting | 67.34 | Semantic Formatting Score | 70.8 |
| AA CritPt | 4 | Accuracy (%) | 67.1 |
| AA GDP.pdf - STEM/Research - Criterion Pass Rate | 68.77 | Criterion Pass Rate (%) | 63.6 |
| AA Humanity's Last Exam | 18.4 | Accuracy (%) | 62.7 |
Interactive version: theaggregate.ai/model?slug=gpt-6-sol-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-12.