Sam Altman said OpenAI “screwed up” GPT-5.2’s writing quality during a developer town hall Monday evening.
When asked about user feedback that GPT-5.2 produces writing that’s “unwieldy” and “hard to read” compared to GPT-4.5, Altman was blunt.
He said:
“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 explained that OpenAI made a deliberate choice to focus GPT-5.2’s development on technical capabilities:
“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.”
How OpenAI Positioned Each Model
The contrast between GPT-4.5 and GPT-5.2 shows where OpenAI focused its resources.
When OpenAI introduced GPT-4.5 in February 2025, the company emphasized natural interaction and writing. OpenAI said interacting with GPT-4.5 “feels more natural” and called it “useful for tasks like improving writing.”
GPT-5.2’s announcement took a different direction. OpenAI positioned it as the most capable model series yet for professional knowledge work, with improvements in creating spreadsheets, building presentations, writing code, and handling complex, multi-step projects.
The release post spotlights spreadsheets, presentations, tool use, and coding. Writing appears more briefly, with technical writing noted as an 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.
Why This Matters
We’ve covered the iterative changes to ChatGPT since GPT-5 launched in August, including updates to warmth and tone and the GPT-5.1 instruction-following improvements. OpenAI regularly adjusts model behavior 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.
If you rely on ChatGPT for writing, treat model updates like any other dependency change. Re-test your prompts when defaults change, and keep a fallback if output quality matters for your workflow.
Looking Ahead
Altman said he believes “the future is mostly going to be about very good general purpose models” and that even coding-focused models should “write well, too.”
No timeline was given for when GPT-5.x writing improvements will ship. OpenAI typically iterates on model behavior through point releases, so changes could arrive gradually rather than in a single update.
Hear Altman’s full statement in the video below:
Featured Image: FotoField/Shutterstock
