GPT-5.6 Pro Sol: benchmark results
OpenAI's GPT-5.6 Sol Pro (July 9, 2026), GPT-5.6 Sol run in its highest-effort Pro reasoning mode for long-running coding, research and cybersecurity work. Provider: OpenAI. Released 2026-07-09. Access: API.
Aggregate Rating 1890 ± 14, rank #18 of 1656 rated models, from 62 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ChessBench GitHub - Elo | 2660 | Benchmark Elo | 100 |
| PrinzBench | 91 | Score (x/99) | 100 |
| Reconstruction (Research Idea Recovery) - Chemistry | 15 | Match rate (%; the 105 chemistry seed papers; share of a mod | 100 |
| Reconstruction (Research Idea Recovery) - Medicine | 14.6 | Match rate (%; the 78 medicine seed papers; share of a model | 100 |
| ReLE - Education - Primary School Subjects | 71.3 | Accuracy (%) | 98.7 |
| Chess Puzzles (Epoch AI) | 68 | Accuracy (%) | 98.4 |
| AIMultiple - FinanceReasoning (Hard) | 90.76 | Accuracy (%, 238 hard FinanceReasoning questions) | 97.2 |
| ReLE - Coding | 75 | Accuracy (%) | 96.7 |
| Arabic Broad Leaderboard - RAG QA | 9.46 | Average Score (0-10) | 96.2 |
| Arabic Broad Leaderboard - Translation (incl Dialects) | 8.06 | Average Score (0-10) | 96.2 |
| FlavourBench - Pairing | 72.46 | Pairing Family Score (0-100) | 96.2 |
| AI Chess Leaderboard (Reasoning) | 1617 | Elo | 95.3 |
Interactive version: theaggregate.ai/model?slug=gpt-5-6-pro-sol · How It Works · Data refreshed daily, snapshot 2026-10-20.