OpenAI has started rolling out GPT 5.2, its latest flagship model for professional work, long documents, and agent style automation. The update brings noticeable improvements in reasoning, coding, accuracy, and vision. It is now available to paid ChatGPT users and through the API.
GPT 5.2 comes in three versions. Instant is the fast option for everyday use. Thinking handles complex tasks and long contexts. Pro is the top tier model for advanced technical and scientific work. All three versions show gains, but Thinking and Pro deliver the biggest jump in quality.
OpenAI says GPT 5.2 Thinking now performs at the level of human experts in most tasks in its GDPval benchmark. It can also generate structured outputs, such as spreadsheets, presentations, and schedules, faster and at a lower cost than manual work.
Coding gets one of the biggest boosts. GPT 5.2 posts new high scores on SWE Bench tests that measure real world software engineering tasks.
The model also improves factual accuracy and reduces incorrect answers compared to GPT 5.1. Vision tasks see upgrades as well. GPT 5.2 is better at reading charts, interpreting UI screenshots, and understanding layouts.
Also Read: Adobe brings Photoshop, Express and Acrobat to ChatGPT
The rollout includes updated safety systems that reduce harmful responses and improve handling of sensitive topics. GPT 5.1 will remain available for three months before it is phased out.
