GPT-5.1
OpenAI · 2025-11-13
GPT-5.1 is OpenAI's API model in the GPT-5 series, designed to balance intelligence and speed across agentic and coding tasks. It dynamically adapts its thinking time to task complexity and supports configurable reasoning effort from none through high.
Benchmark scores
| Benchmark | Score |
|---|---|
| A-Fantasia | 51.0 |
| A-Fantasia Backwards Spelling | 63.0 |
| A-Fantasia Chess | 16.0 |
| A-Fantasia Cube Rotation | 74.0 |
| AA-LCR | 75.0 |
| AA-Omniscience Index | 5.5 |
| Agent Red Teaming k=100 (CAIS) | 82.1 |
| AHK-Eval | 69.4 |
| AHK-Eval Algorithms | 66.7 |
| AHK-Eval Data Structures | 33.3 |
| AHK-Eval Date and Time | 33.3 |
| AHK-Eval Easy | 50.0 |
| AHK-Eval Hard | 58.3 |
| AHK-Eval Hidden Cases | 64.6 |
| AHK-Eval Mid | 75.0 |
| AHK-Eval Numbers | 100.0 |
| AHK-Eval Parse Success | 100.0 |
| AHK-Eval Regex | 50.0 |
| AHK-Eval Strings | 83.3 |
| AIME 2024 (avg@8) | 93.8 |
| AIME 2024-2025 | 93.3 |
| AIME 2025 | 38.0 |
| AIME 2025 (avg@8) | 92.9 |
| ALE-Bench | 1425 |
| Apex Shortlist | 55.9 |
| APEX-Agents | 17.5 |
| ARC-AGI-1 | 33.2 |
| ARC-AGI-2 | 6.5 |
| Arena Code WebDev — Brand & Marketing | 1394 |
| Arena Code WebDev — Consumer Product | 1399 |