GPT-5.6 Sol (Low): benchmark results
GPT-5.6 Sol evaluated at the low reasoning-effort setting. Provider: OpenAI. Released 2026-07-09. Access: API.
Aggregate Rating 1786 ± 12, rank #166 of 2168 rated models, from 139 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| FinLifeBench - Financial State - Evidence Recall | 88.1 | Recall of gold evidence sessions (%) | 100 |
| Taiwan Exams - Counseling Psychologist Exam (ROC 115) | 97.83 | Points Scored (%) | 100 |
| AA Terminal-Bench Hard | 60.61 | Accuracy (%) | 98.5 |
| AA-Omniscience Index - Software Engineering (SWE) - PHP | 84 | Omniscience Index | 97.6 |
| AA-Omniscience Index - Software Engineering (SWE) - JavaScript | 78.18 | Omniscience Index | 96.2 |
| AA-Omniscience Index - Software Engineering (SWE) - Kotlin | 76 | Omniscience Index | 95.7 |
| AA-Omniscience Accuracy | 57.17 | Accuracy (%) | 95 |
| AA-Omniscience Index - Software Engineering (SWE) - TypeScript | 80 | Omniscience Index | 94.1 |
| AA-Omniscience Index - Software Engineering (SWE) - Go | 72 | Omniscience Index | 93.7 |
| AA-Omniscience Index - Software Engineering (SWE) - Swift | 68 | Omniscience Index | 93.6 |
| AA-Omniscience Index - Software Engineering (SWE) - C | 80 | Omniscience Index | 93.5 |
| AA-Omniscience Index - Software Engineering (SWE) | 66.5 | Omniscience Index | 93.4 |
Interactive version: theaggregate.ai/model?slug=gpt-5-6-sol-low · How It Works · Data refreshed daily, snapshot 2026-10-20.