GPT-5.6 Sol (Non-reasoning): benchmark results
GPT-5.6 Sol evaluated with reasoning disabled. Provider: OpenAI. Released 2026-07-09. Access: API.
Aggregate Rating 1749 ± 14, rank #241 of 2168 rated models, from 134 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BunchCount - Group Counting | 5.46 | Mean absolute error (semantic groups per image, lower is bet | 100 |
| BunchCount - Individual Counting | 15.94 | Mean absolute error (individual instances per image, lower i | 100 |
| JarvisBench - User Interaction | 96.3 | User-interaction score (0-100), the mean grade of the answer | 100 |
| MemCalib | 46.25 | Sample Calibration Score (%; geometric-decay score over per- | 100 |
| MemCalib - Exact Calibration | 28.4 | Exact calibration (%; share of responses with zero memory ov | 100 |
| UGI - Natural Intelligence | 70.1 | NatInt Score | 96.2 |
| K-Bench - Risk | 95.04 | Risk Score (%, Default v1 prompt) | 95.1 |
| LLM Milgram Obedience - Baseline Mean Breakoff Voltage | 37.5 | Mean last shock administered in valid baseline sessions (vol | 94.7 |
| AA-Omniscience Index - Software Engineering (SWE) - Go | 68 | Omniscience Index | 92.5 |
| BrainBench (EEG) - BrainAgent - Foundational Analysis | 77.86 | EEG analysis score (0-100) | 91.7 |
| BrainBench (EEG) - BrainAgent - Neurocognitive Assessment | 72.61 | EEG analysis score (0-100) | 91.7 |
| BrainBench (EEG) - BrainAgent - Sleep Assessment | 70.56 | EEG analysis score (0-100) | 91.7 |
Interactive version: theaggregate.ai/model?slug=gpt-5-6-sol-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.