I put GPT-5.5 through a 10-round test: It scored 93/100, losing points only for exuberance (www.zdnet.com)

OpenAI has unveiled GPT-5.5, a noteworthy upgrade from its predecessor GPT-5.4, boasting enhanced performance in writing, coding, and reasoning tasks. A rigorous 10-round testing process revealed that GPT-5.5 scored 93 out of 100, with its main shortfall being an overzealous approach that occasionally resulted in the model not following specific instructions accurately. For instance, while it provided comprehensive summaries, it often drew from sources beyond the requested one, demonstrating a potential risk in deploying AI for autonomous tasks. The faster release cadence of AI improvements is largely attributed to advancements in coding efficiency, driven by the capabilities of AI itself. Notable strengths of GPT-5.5 included its ability to explain complex concepts in accessible language, perform mathematical analysis accurately, and engage effectively in creative writing, producing a lengthy story that captivated the tester. Its performance underscores progress in natural language understanding and application, setting the stage for more sophisticated AI interactions, while also highlighting the importance of precision in instructional adherence as AI systems become increasingly autonomous.