- The AI Daily Brief
- Posts
- GPT-5.2 is Here and It's For Professionals, Dammit
GPT-5.2 is Here and It's For Professionals, Dammit
Did OpenAI rise to the Code Red challenge?
Nathaniel Whittemore
December 11, 2025
We got the not-so-long-because-we-didn’t-even-know-it-was-a-thing-until-last-week-awaited new OpenAI model GPT-5.2 and I’m not sure I’ve ever seen OpenAI staffers so consistence: this is a model for professionals to do valuable, real-world, professional things.
The benchmarks OpenAI shared are impressive - spectacularly so in some cases (I’m thinking about ARC-AGI-1, where 5.2 demonstrates that AI has gotten 390x more efficient in a single year 🤯).
But the one they chose to highlight was GPDval, their recently-introduced measure of AI’s ability to complete economically relevant knowledge work tasks across 44 professions. GPT-5.2 thwumped GPT-5 in that category, basically doubling its score.
By all accounts the model is massively better on spreadsheets and powerpoints and truly you’ve never seen a company so excited about spreadsheets and powerpoints. Which makes sense given that combined they constitute approximately 97.3% of modern “work.”
People’s first impressions are positive. Some think the progress is incremental. Others think it’s quite good, but slow in a way that only makes it appropriate for really heavy thinking tasks. Others think that even though it’s slow, it’s so good that it’s indispensable. It’s rolling out to everyone now so we’ll all get to decide for ourselves quite soon.
In the meantime, the thing I’m most keeping an eye on is the shift in long-context capabilities. This is a fairly massive change with some big implications for many important use cases.
Have you tried 5.2 yet? What do you think?
P.S. as if that weren’t enough, OpenAI also announced a massive deal with Disney that will see 200 characters come to Sora, Sora content using those characters come to Disney+, and Disney become a major customer and $1B investor in OpenAI.
