January 29, 2026 OpenAI knowingly traded writing quality for technical gains in GPT-5.2 and got that balance wrong, CEO Sam Altman says. According to the CEO, the company is working on improvements for future GPT-5.x models.
Speaking during a developer town hall on Monday, Sam Altman acknowledged that GPT-5.2’s writing output has frustrated users and lagged behind earlier models, particularly GPT-4.5.
“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.”
The comment follows weeks of user feedback describing GPT-5.2’s prose as harder to read and less fluid than previous releases. According to Altman, the issue was the result of a deliberate prioritization decision.
“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.”
“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.”
The contrast between GPT-4.5 and GPT-5.2 reflects how OpenAI has positioned each release. When GPT-4.5 launched in February 2025, OpenAI highlighted conversational fluency and writing quality. The company said interacting with the model “feels more natural” and described it as “useful for tasks like improving writing.”
GPT-5.2 was framed very differently. Its release emphasized professional and technical work: spreadsheets, presentations, tool use, code generation, and multi-step problem solving. Writing appeared mainly in a technical context, with OpenAI noting improvements to technical documentation in GPT-5.2 Instant rather than broad narrative or editorial prose.
OpenAI has made frequent, incremental changes to ChatGPT since GPT-5 launched in August, including updates to tone, warmth and instruction-following in GPT-5.1.
Altman said he expects future versions to close the gap. He added that he believes “the future is mostly going to be about very good general purpose models” and that even models optimized for coding and reasoning should “write well, too.”
No timeline was given for when writing improvements will land in GPT-5.x. Based on OpenAI’s past pattern, changes are more likely to arrive gradually through point releases rather than as a single reset.
