I tested GPT-5.4 Thinking, and it gave me great answers (until I dove deeper)

Editor's take

OpenAI's internal testing of a new model, reportedly dubbed GPT-5.4 Thinking, revealed its capabilities in professional tasks, though limitations emerged upon deeper scrutiny. This development is significant as it signals OpenAI's ongoing efforts to achieve higher levels of reasoning and task completion, a critical step towards more general-purpose AI applications that could impact various industries from software development to scientific research.

The key question is whether this iteration truly advances beyond the current state-of-the-art, such as GPT-4, in terms of reliable, nuanced problem-solving. Future performance benchmarks and independent evaluations will be crucial to assess if the perceived limitations are inherent to this specific model or indicative of broader challenges in achieving robust AI reasoning.