Saad Ullah
⬤ TestingCatalog News reported a key change in GPT-5.1: the model now respects custom instructions about em-dash usage. This comes alongside broader improvements in how well the model follows directions compared to GPT-5, marking a shift toward more reliable and user-controlled behavior.
⬤ OpenAI confirmed that GPT-5.1 handles em-dash preferences better when set through custom instructions. This isn't just a side effect—it's a deliberate improvement meant to give users more precise control over formatting. For writers and developers who need consistent style compliance, this matters.
⬤ The real story here is GPT-5.1's improved ability to stick to specific, detailed instructions. Earlier models would sometimes drift from user-defined rules during longer tasks. By fixing issues like unwanted formatting, GPT-5.1 shows OpenAI is working to make the model behave more like users actually expect it to.
⬤ Early feedback suggests this release focuses on refining everyday usability rather than adding flashy new features. Better instruction following and predictable formatting make GPT-5.1 a solid upgrade for anyone working with detailed content or automated workflows.
Saad Ullah
