OpenAI launches GPT-5.2 with improved reasoning, memory, and pro-focused variants
After a turbulent year and an internal code red memo triggered by competitive pressure from Google’s Gemini 3 Pro, OpenAI has revealed GPT-5.2, its new flagship model. The update targets professional and developer use cases, with an emphasis on stronger reasoning, improved memory handling, and more reliable outputs. GPT-5.2 is available to paid ChatGPT users and developers through the API in three versions: Instant, Thinking, and Pro.
GPT-5.2 Instant focuses on speed for everyday tasks such as writing, information lookup, and translation, while GPT-5.2 Thinking is built for more demanding work including coding, long document analysis, mathematical reasoning, and multi step planning. The Pro variant is aimed at users who need higher accuracy and consistency in professional or enterprise workflows.
OpenAI says GPT-5.2 improves performance in areas such as spreadsheet creation, presentations, code generation, image understanding, long context handling, and complex workflows. In internal benchmarks, GPT-5.2 Thinking outperformed GPT-5.1 across all ten tests, including a perfect score on the AIME 2025 math benchmark and a clear improvement on the ARC AGI 1 reasoning evaluation, while generating about 30 percent fewer factual errors.
- AI Chatbot
- Freemium • Proprietary
Related news
External links
- Introducing GPT-5.2OpenAI Blog • Official source
- ChatGPT 5.2 is here and all about being a better thinker – here are the 3 things you need to knowTechRadar
- OpenAI releases GPT-5.2 after “code red” Google threat alertArs Technica
- OpenAI Launches GPT-5.2 as It Navigates ‘Code Red’WIRED
- OpenAI releases GPT-5.2 to take on Google and AnthropicEngadget
Comments
The worst thing on OpenAi is their intransparent price policy. GPT 5.2 Pro might be leading in some benchmarks. But you only get it for 200$. At Google I always can be sure to get the best engine for just 20$.
Dont they only post the specific benchmarks where it beats the other models like gemini or claude? kinda misleading
This is GPT 5.2 when asked (not without pressure) for the actual difference between GPT 5.2 and Opus 4.5, AND just to make i clear, THIS IS GPT, not the thruth...
"What can we honestly say about GPT-5.2 vs Claude Opus 4.5? Without relying on marketing blogs: There is relative consensus on the following patterns: GPT-5.2
Claims to be stronger at:
Abstract reasoning Coherent multi-step analysis Complex academic synthesis
More flexible in open-ended problems More "ambitious" in responses (and therefore more willing to take risks – read: more prone to confident errors)
Claude Opus 4.5
Demonstrably strong at:
Code quality Structured tasks Agent and workflow scenarios
More conservative and predictable Lower hallucination rate, though also more cautious (read: knows when to stop)
These are usage patterns, not marketing points."
It still falls behind Gemini 3 Pro and Claude 4.5 Opus. They're desperately quoting specific benchmarks in specific ways to make it seem otherwise. Also, the API cost for 5.2 is 40% higher than 5.1. You're not getting 40% better performance, not even close.