OpenAI CEO Sam Altman has revealed a significant mistake that the company has made to its developers. During a recent town hall meeting, Altman acknowledged an issue in GPT-5.2’s writing quality. He cited user complaints to disclose that the latest model produces text that’s “unwieldy” and “hard to read” compared to GPT-4.5.“I think we just scr**ed that up. We will make future versions of GPT 5.x hopefully much better at writing than 4.5 was,” Altman said when asked about the feedback. He admitted that OpenAI also deliberately chose to focus ChatGPT’s GPT-5.2 model’s development on technical capabilities and neglect its writing skills.“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,” Altman explained. The admission comes as OpenAI faces competition in the AI space, and user expectations around model performance continue to evolve.
How ChatGPT models’ writing is expected to improve in the coming days
The contrast between GPT-4.5 and GPT-5.2’s writing skills shows where OpenAI focused its resources. When OpenAI introduced GPT-4.5 in February 2025, the company emphasised natural interaction and writing. The company said interacting with GPT-4.5 “feels more natural” and called it “useful for tasks like improving writing”.Meanwhile, GPT-5.2’s announcement took a different direction. OpenAI positioned it as a capable model series for professional knowledge work, with improvements in spreadsheet creation, presentation design, and code development, as well as in handling complex, multi-step projects.The model’s launch announcement focused on spreadsheets, presentations, tool use, and coding. Writing was less emphasised, with technical writing identified as an area for improvement for GPT-5.2 Instant. But Altman’s comments suggest the overall writing experience still fell short for users, comparing it to GPT-4.5.ChatGPT has undergone numerous updates since GPT-5 launched in August 2025, including improvements to warmth and tone in the GPT-5 instruction followers. OpenAI regularly adjusts model behaviour based on user feedback, and regressions in one area while improving another aren’t new.What’s unusual is hearing Altman acknowledge a tradeoff this directly. For anyone using ChatGPT output in client-facing work, drafts, or polished writing, this explains why outputs may have changed. Model upgrades don’t guarantee improvement across every capability.For those who rely on ChatGPT for writing, treating model updates like any other dependency change may be helpful. Retesting prompts when defaults change, and maintaining a fallback if output quality matters for workflows, could help ensure consistency.Altman also said that he believes “the future is mostly going to be about perfect general-purpose models” and that even coding-focused models should “write well, too.”He didn’t provide a timeline for when the ChatGPT model’s writing improvements will be rolled out. OpenAI typically iterates on model behaviour through point releases, so changes arrive gradually rather than in a single update.
