Article
December 11, 2025
Introducing GPT-5.2
# Work Users
The latest upgrade to the GPT-5 series
Introducing GPT-5.2 for Work
GPT-5.2 is the newest upgrade in the GPT-5 family and our smartest, most capable model series yet for work and learning—while remaining just as enjoyable and intuitive to use. As with GPT-5.1, this release reflects our approach of making regular, meaningful improvements to our core models so they become more helpful over time without regressing on speed, tone, or usability.
All GPT-5.2 models now share a new August 2025 knowledge cutoff, giving the model a more current understanding of the world from the start. That means more accurate answers, more relevant examples, and less time spent correcting or filling in gaps.
What’s better in GPT-5.2?
GPT‑5.2 brings a step‑up in intelligence and reliability across everyday tasks and complex, professional workflows.
- More up‑to‑date knowledge: With an August 2025 knowledge cutoff, GPT‑5.2 starts with stronger context for current events, tools, and trends—so responses are clearer and more useful right away.
- Stronger everyday performance with GPT‑5.2- Instant: GPT‑5.2 Instant is better at everyday work and learning tasks, producing clearer, more comprehensive, and immediately usable outputs. You’ll notice improvements in writing, translation, procedural guidance, and general information‑seeking.
- Deeper reasoning with GPT‑5.2 -Thinking: GPT‑5.2 Thinking, our model built for complex professional queries, achieves new highs across math, science, coding, multimodal, and long‑context benchmarks. It’s especially well‑suited for multi‑step analysis, planning, and decision‑making where structure and accuracy matter.
- Better artifact creation: Compared to GPT‑5.1, GPT‑5.2 shows improvements in creating and editing slides and spreadsheets, producing outputs that are more structured, scannable, and ready to use.
How to get started
GPT‑5.2 will soon be the default experience in ChatGPT, but we're releasing GPT-5.2 in early access for enterprises to enable early for their teams.
In most cases, you don’t need to think about models at all—the system automatically routes each prompt to the model (Instant or Thinking) best suited for the task. If you prefer manual control, you can also choose a specific model:
- GPT-5.2-Instant: A fast, everyday workhorse for work and learning, delivering clearer, more upfront explanations for info-seeking, how-tos, technical writing, and translation with a warm conversational tone.
- GPT-5.2-Thinking: Built for deeper, more complex work, offering greater polish and structure for coding, long-document summarization, file analysis, step-by-step math and logic, and planning and decision support.
- GPT-5.2-Pro: The most intelligent and reliable option for hard problems where quality matters more than speed, with fewer major errors and stronger performance in complex domains like programming.
Please note that Custom GPTs will be transitioning to using GPT-5.2 on January 12, 2025. We strongly recommend testing and moving custom GPTs to GPT-5.2 as soon as possible.
Try GPT-5.2 out for yourself:
One of the best ways to experience what’s new is to try GPT‑5.2 across a mix of simple and complex tasks.
Use GPT‑5.2 Instant for quick work that doesn’t require deep reasoning.
Write a short email (5–7 sentences) to my team explaining [topic]. The tone should be [tone]. Include why this matters, what’s changing, and what they need to do next.
Use GPT‑5.2-Thinking when tasks are ambiguous, high‑stakes, or multi‑step.
Act as a senior strategy and communication partner. Review the attached announcement, identify risks or gaps, anticipate stakeholder concerns, and propose a clear communication plan. Rewrite the announcement using plain language and clear next steps.
Key actions for admins
- Test out GPT-5.2 vs previous models This is a great time to encourage your team to test out the new model and observe any new behaviors. Use this prompt (replace your company name) to generate a table of use cases to test out with GPT-5.2.
- Review and test custom GPTs We’ll be automatically updating all custom GPTs to GPT-5.2 on January 12, 2025. We strongly recommend testing and moving custom GPTs to GPT-5.2 as the recommended model as soon as possible. We especially recommend testing business-critical custom GPTs with Actions.
- Share the announcement with your team: This is a great time to explore new use cases for work, refresh training on best practices, and encourage your team to experiment. Share resources on the OpenAI Academy for Work, and use the sample messaging below to explain the model change:
Hey team! You may notice that your default model in ChatGPT has been updated to GPT-5.2 this week. GPT-5.2 is a major upgrade for work: it’s smarter and more capable at complex, end-to-end tasks—delivering more accurate, usable outputs with fewer mistakes. With an updated August 2025 knowledge cutoff across all models, it also starts with more current context, so you spend less time correcting or filling gaps.
[Add in details about accessing legacy models, if available]
To test it out, I recommend you:
- Try out some more complex strategic questions to see the reasoning power of GPT-5.2-Thinking.
- Review any custom GPTs you have built, and test them with GPT-5.2. We recommend switching the Recommended Model ASAP, as these will automatically migrate in January 2026.
- Share new use cases you’ve been able to unlock with your team!
- Manage access to legacy models While the default experience will be GPT-5.2, you can still manage access to legacy models for your team. See details about legacy model access in our Help Center.
Additional resources
Table Of Contents