OpenAI CEO Sam Altman has candidly admitted that the company "screwed up" the writing quality of its latest large language model, GPT-5.2. During a recent developer town hall, Altman explained that a deliberate focus on enhancing the model's coding, reasoning, and engineering capabilities led to a noticeable decline in its ability to generate natural, readable text, a feature users had praised in earlier versions like GPT-4.5. He assured the community that future GPT-5.x iterations would prioritize and significantly improve writing proficiency.

Addressing user feedback that described GPT-5.2's output as "unwieldy" and "hard to read" compared to its predecessor, GPT-4.5, Altman was direct in his response.

"I think we just screwed that up. We will make future versions of GPT 5.x hopefully much better at writing than 4.5 was," Altman stated.

He further elaborated on the strategic decision behind GPT-5.2's development focus.

"We did decide, and I think for good reason, to put most of our effort in 5.2 into making it super good at intelligence, reasoning, coding, engineering, that kind of thing. And we have limited bandwidth here, and sometimes we focus on one thing and neglect another," he explained.

OpenAI's Shifting Model Focus

This contrast in performance highlights OpenAI's evolving priorities across its generative AI models.

When OpenAI introduced GPT-4.5, the company heavily emphasized its natural interaction and writing prowess, marketing it as a tool that "feels more natural" and "useful for tasks like improving writing."

However, the announcement for GPT-5.2 pivoted significantly. This model was positioned as OpenAI's most capable series yet for professional knowledge work, with key improvements highlighted in areas such as creating spreadsheets, building presentations, writing code, and managing complex, multi-step projects. While technical writing was briefly noted as an improvement for GPT-5.2 Instant, Altman's recent comments confirm that the overall user experience for general writing tasks fell short when compared to GPT-4.5.

Implications for AI Users and Developers

OpenAI frequently refines its ChatGPT models based on user feedback, and it's not uncommon for improvements in one area to coincide with regressions in another. The company has, for instance, made iterative changes to ChatGPT since the GPT-5 launch in August, including updates to warmth, tone, and instruction-following.

What makes Altman's recent acknowledgment particularly noteworthy is its directness. For professionals who integrate ChatGPT's output into client-facing work, drafts, or polished content, this explanation clarifies why the quality of generated text may have varied. It serves as a crucial reminder that model upgrades do not automatically guarantee universal improvement across every capability.

Users who rely on ChatGPT for writing tasks are advised to treat model updates like any other dependency change in their workflow. It's prudent to re-test prompts when default behaviors shift and to maintain fallback strategies if the quality of output is critical to their operations.

The Path Ahead for GPT-5.x

Looking to the future, Altman expressed his belief that "the future is mostly going to be about very good general purpose models" and that even models primarily focused on coding should "write well, too."

While no specific timeline was provided for the implementation of GPT-5.x writing improvements, OpenAI typically rolls out model behavior adjustments through point releases. This suggests that enhancements to writing quality are likely to arrive gradually rather than in a single, comprehensive update.

For a full account of Altman's statement, you can watch the video below:

Featured Image: FotoField/Shutterstock