- The new ChatGPT 5.2 AI model delivers strong performance in professional knowledge work tasks like creating spreadsheets and building presentations.
- GPT-5.2 Thinking is also very capable at coding, understanding long contexts, analyzing visual data, using multiple tools, and more.
- ChatGPT 5.2 is rolling out today to paid users including Go, Plus, Pro, Business, and Enterprise.
OpenAI has finally released GPT-5.2, its most advanced frontier AI model yet and it’s already rolling out to ChatGPT users. After declaring a “code red” situation internally, the new ChatGPT 5.2 model is seen as OpenAI’s direct answer to Google’s Gemini 3 Pro. The new ChatGPT 5.2 update brings three new AI models: GPT-5.2 Instant, GPT-5.2 Thinking, and GPT-5.2 Pro.
What is New in ChatGPT 5.2?
OpenAI says the new ChatGPT 5.2 models are highly advanced for “professional work and long-running agents.” For professional knowledge work, you will find that GPT-5.2 delivers the best results. ChatGPT 5.2 is currently state-of-the-art for writing code, creating spreadsheets, building presentations, understanding long context information, analyzing visual data, using tools and more.
GPT-5.2 can also handle complex, multi-step projects and use multiple tools without confusion. As for benchmarks, the new ChatGPT 5.2 model hits it out of the park and beats both Gemini 3 Pro and Claude Opus 4.5 across many benchmarks.
In the SWE-Bench Pro benchmark which measures performance on software engineering tasks, GPT-5.2 Thinking achieved 55.6%, higher than Claude Opus 4.5 (52.0%) and Gemini 3 Pro (43.3%). Next, in SWE-bench Verified, GPT-5.2 Thinking got 80.0%, but Claude Opus 4.5 continued to maintain its supremacy with 80.9%.
In GPQA Diamond (a benchmark consisting science questions), GPT-5.2 Thinking achieved 92.4% without tool use, again beating Gemini 3 Pro. Notably, in ARC-AGI-2, GPT-5.2 Thinking scored 52.9%, much higher than Claude Opus 4.5 (37.6%) and Gemini 3 Pro (31.1%).
And on GDPval — a benchmark measuring well-specified knowledge work across 44 occupations — GPT-5.2 Thinking matches or performs better than top industry professionals on 70.9% of tasks.
ChatGPT 5.2 is rolling out today to paid users including Go, Plus, Pro, Business, and Enterprise. The knowledge cutoff date of GPT-5.2 is August 2025 which means it’s a fairly new AI model. Meanwhile, OpenAI also suggested that adult mode in ChatGPT is coming in the first quarter of 2026. OpenAI is testing an age prediction model for users under 18.
