On this page
No sooner had I been speaking this week about Gemini's current lead in areas like image generation, multimodal document review, and large-context reasoning… OpenAI released GPT-5.2 today.
And it's a meaningful step forward.
What's New in GPT-5.2
In simple terms, GPT-5.2 is noticeably stronger at reasoning, decision-making, and complex, multi-step tasks. It makes fewer errors, handles larger documents more reliably, and shows a sharp lift in performance on tasks that mirror real knowledge work.
One example is the GDPval benchmark: a test where models complete tasks that would normally take a human specialist 4–8 hours. GPT-5.2 now wins around 71% of the time, up from under 50%. That's a significant shift in only a few months.
Different Tools for Different Jobs
For my own work, I continue to use both models:
- OpenAI for deep reasoning, research, contract analysis, and multi-step thinking
- Gemini for image-heavy work, video production, multimodal scans, and very large document contexts
Different tools for different jobs, and both are evolving at a pace that's hard to overstate.
Why This Matters
Today is another reminder of how fast this space is moving, and why it pays to have someone you trust keeping an eye on the landscape, helping you choose the right model for the right task, and ensuring your workflows can adapt as the technology leaps forward. For the longer view on how to think about model selection structurally, see The Model Wars: Don't Get Caught on the Wrong Side.
