GPT-5.2 Has Arrived. Practical Improvements for Real Work
I wanted to share a quick update thatโs worth paying attention to if you use ChatGPT regularly. OpenAI has just released GPT-5.2, and itโs now live in ChatGPT as well as the API. A few highlights that stood out to me ๐ โข Much stronger for real workGPT-5.2 Thinking is the first model to hit expert-level scores on OpenAIโs GDPval benchmark, which tests real professional tasks like writing documents, building spreadsheets, creating presentations, and problem-solving. โข Fewer mistakesCompared to GPT-5.1, responses with errors were reduced by around 38%. That matters if youโre using ChatGPT for research, planning, writing, or decision-making rather than just casual questions. โข Clearer model choicesYouโll now see three main options: โ Instant for fast answers, how-tos, and learningโ Thinking for deeper reasoning, structured work, and complex tasksโ Pro when accuracy matters more than speed โข Big improvements for developers and agentsIf you build tools or workflows with AI, GPT-5.2 is much better at long-context understanding, tool calling, and completing multi-step tasks end-to-end. Overall, this feels like a meaningful step forward for people actually using ChatGPT to get things done, not just experiment. If youโve already noticed differences when using ChatGPT today, Iโd love to hear what youโre seeing.What feels better? What still feels the same? Cheers Jason