GPT-5.6 Sol Pro
OpenAI · 2026-07-09
GPT-5.6 Sol Pro is the higher-compute Pro variant of GPT-5.6 Sol, represented as a distinct Atlas model so benchmark results reported under the Pro identity are not collapsed into standard Sol. OpenAI exposes Pro through the gpt-5.6-sol API model with reasoning.mode set to pro rather than through a separate API model ID.
Benchmark scores
| Benchmark | Score |
|---|---|
| AA-LCR v1.1 | 85.3 |
| AdvancedIF | 82.1 |
| AdvancedIF (rubric mean score) | 94.6 |
| AHK-Eval | 97.2 |
| AHK-Eval Algorithms | 100.0 |
| AHK-Eval Data Structures | 100.0 |
| AHK-Eval Date and Time | 100.0 |
| AHK-Eval Easy | 91.7 |
| AHK-Eval Hard | 100.0 |
| AHK-Eval Hidden Cases | 97.8 |
| AHK-Eval Mid | 100.0 |
| AHK-Eval Numbers | 100.0 |
| AHK-Eval Parse Success | 100.0 |
| AHK-Eval Regex | 83.3 |
| AHK-Eval Strings | 100.0 |
| AIME 2025 | 100.0 |
| APEX (Mercor) | 22.1 |
| APEX-Accounting | 51.5 |
| APEX-Accounting Pass@1 | 9.6 |
| APEX-Agents | 40.0 |
| APEX-Agents 1.1 | 65.2 |
| APEX-Agents 1.1 Pass@1 | 51.5 |
| APEX-v1-extended | 73.1 |
| BrowseComp (Mercor 130-question subset) | 95.7 |
| BuseyBench SVG | 6.3 |
| CadQueryEval | 84.0 |
| CadQueryEval — Bounding Box | 100.0 |
| CadQueryEval — Chamfer Distance | 100.0 |
| CadQueryEval — Code Execution | 100.0 |
| CadQueryEval — Component Count | 100.0 |