GPT 5.4 is here but Sam Altman still thinks GPT 5.3 is better, here is why
Follow UsGPT 5.4 feels more natural and human-like in conversations.
Sam Altman says GPT 5.3 is still strong for coding and tasks.
Users may choose based on conversation vs performance needs.
OpenAI launched GPT 5.4 at the start of the month, and many users have praised its abilities. Yet in a surprising turn of events, CEO Sam Altman shared a different view. He said that GPT 5.3 still has strengths that developers and users value. Replying to a post on X, Altman noted GPT 5.4 for its human-like responses and personality. At the same time, he agreed with the post author that the older model remains impressive in key areas. His comment suggests that progress does not always replace what worked well before and that both versions can serve different needs for people, as many teams may pick a model based on their specific tasks.
SurveyWhy GPT 5.4 is getting attention
The latest model from OpenAI is already seeing rapid adoption. Company president Greg Brockman recently said that GPT 5.4 usage has grown faster than any previous model in the company’s history. Within a week of launch, developers were processing trillions of tokens every day through the API.
The most significant change that users notice is how natural the model feels in conversations. The answers it gives seem more thoughtful and expressive. It does not just provide answers in a robotic fashion but in a way that feels more like a human would talk. This approach is why a number of developers and creators online say that GPT 5.4 feels more engaging as a model for use.
Also read: AI may not take your jobs but may lead to salary cut, says former SalesForce CEO
Why Sam Altman still values GPT 5.3
Despite the excitement, Altman suggested that GPT 5.3 still stands out in certain areas. In his reply on X, he said the new version is very smart, but he personally felt the difference when moving from 5.3 to 5.4.
Some developers believe GPT 5.3 remains extremely strong when it comes to structured tasks such as coding and technical problem-solving. Earlier feedback from programmers often described GPT 5.3 as highly reliable when writing or fixing code.
Also read: Samsung Galaxy Z Fold 8 leaks: Expected features, price and launch details
As a result, many people continue to compare the two versions depending on what they need. For some users, personality and conversation matter more. For others, accuracy and task-focused performance remain the top priority.
