OpenAI Rolls Out GPT-5.2 With Expanded Capabilities for ChatGPT Users

OpenAI introduced GPT-5.2 on December 11, 2025, as a new series of artificial intelligence models designed to enhance the performance of its ChatGPT service. The release features three variants, GPT-5.2 Instant, GPT-5.2 Thinking, and GPT-5.2 Pro, each optimized for specific applications in professional work.

GPT-5.2 Instant focuses on quick responses for routine tasks, while GPT-5.2 Thinking handles coding, math, and planning with step-by-step logic, and GPT-5.2 Pro addresses complex queries with higher accuracy.

The models incorporate a 400,000-token context window in certain configurations, allowing the processing of large codebases or extensive documents in one interaction.

Developers can access GPT-5.2 through OpenAI's application programming interface immediately, with variants available under names such as gpt-5.2 and gpt-5.2-pro.

In ChatGPT, the rollout began on December 11 for subscribers to paid plans including Plus, Pro, Business, and Enterprise, with phased implementation to ensure stability.

OpenAI trained the models using infrastructure from partners Microsoft and Nvidia, involving Azure data centers and advanced graphics processing units.

GPT-5.2 shows gains in benchmarks for coding, science, math, and reasoning, achieving scores like 55.6 percent on SWE-Bench Pro for software engineering tasks and 92.4 percent on GPQA Diamond for expert-level questions.

The series reduces hallucinations, with GPT-5.2 Thinking showing a 10.9 percent error rate on factual queries, down from previous versions. It also improves handling of spreadsheets, presentations, image perception, and multi-step projects.

"We designed GPT‑5.2 to unlock even more economic value for people; it’s better at creating spreadsheets, building presentations, writing code, perceiving images, understanding long contexts, using tools, and handling complex, multi-step projects." OpenAI stated in an official announcement.

The launch followed an internal "code red" directive from OpenAI CEO Sam Altman in early December, redirecting resources to accelerate development in response to competition from Google's Gemini 3 model.

"Gemini 3 has had less of an impact on our metrics than we feared," said Sam Altman in an interview with CNBC on Thursday, alongside Disney CEO Bob Iger.

OpenAI plans to maintain access to older models like GPT-5.1 in its API without immediate deprecation, and the company intends to release a Codex-optimized version of GPT-5.2 in the coming weeks.

The models include strengthened safety measures for sensitive topics such as mental health and self-harm, resulting in fewer undesirable responses compared to prior iterations.