GPT-5 is OpenAI's flagship text model, known for strong reasoning and coherent long-form generation. It handles complex multi-step problems with notable consistency, though like all large models it can still hallucinate facts confidently. It tends to be thorough, sometimes verbose, when precision is needed.
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| SWE-Bench | 74.9 | accuracy | 5d ago |
| AIME 2025 | 94.6 | accuracy | 5d ago |
| Aider Polyglot | 88.0 | accuracy | 5d ago |