Sam Altman (샘 알트먼), CEO of OpenAI, rated GPT-5.4 as the "most suitable model for conversation" while acknowledging three weaknesses that need improvement.

On March 9, local time, IT outlet TechRadar reported that Altman said, "We have missed the mark on model personality for awhile, so it feels extra good to be moving in the right direction."

His use of the phrase "for awhile" appears to refer to debate over ChatGPT's personality that has troubled OpenAI since it phased out GPT-4o, a popular model, in November 2025. Many users consider GPT-4o to have a much better personality than subsequent fifth-generation GPT models. Altman also strongly praised GPT-5.4's conversational performance, saying earlier models were lacking in terms of interaction with humans. He expressed satisfaction, saying, "GPT-5.4 also shows strong performance in writing code," but pointed out that improvements are still needed in certain functions.

For example, GPT-5.4 is assessed as lacking a sense of visual design when generating user interfaces, and failing to produce UIs as appealing as competing models Claude and Gemini.

Compatibility issues with OpenClaw, which automates artificial intelligence workloads, also remain. OpenClaw is a system that runs AI models using a Mac mini, and critics say GPT-5.4 needs optimization. But Altman promised improvements, saying, "These problems can be solved."

The debate suggests OpenAI is now considering AI user experience as important beyond simple performance. GPT-5.4 has set a new standard for conversational AI, but tasks still remain to make it a complete model.

GPT-5.4 is great at coding, knowledge work, computer use, etc, and it's nice to see how much people are enjoying it. But it's also my favorite model to talk to! We have missed the mark on model personality for awhile, so it feels extra good to be moving in the right direction.