We’re experiencing unreliable behavior from the GPT-4.1 model when used via the OpenAI API, specifically related to its function/tool calling behavior. The model frequently fails to follow well-defined instructions or skips invoking tools even when the schema and prompts are correctly structured.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| New gpt-4-0125-preview model will not use tools unless asked to | 3 | 4690 | February 1, 2024 | |
| GPT-4o with function calling not reliable from Assistants-API | 1 | 125 | August 1, 2025 | |
| Function calling became unreliable | 1 | 559 | October 9, 2024 | |
| Unexpected Tool Call Behavior with Response Format and Tool Descriptions in GPT-4o-mini API | 4 | 490 | December 11, 2024 | |
| Gpt-4o-2024-08-06 not reliably calling functions | 3 | 351 | March 7, 2025 |