ChatGPT 5.2 update is here: List of new capabilities OpenAI's chatbot gets with the new GPT 5.2 model

GPT 5.2

OpenAI has announced GPT-5.2, its most advanced model series for professional and long-running agent workflows. The update is now rolling out to paid ChatGPT users, including Plus, Pro, Business, and Enterprise plans. GPT-5.2 brings improvements across reasoning, factual accuracy, long-context processing, coding capability, and vision performance, making ChatGPT more reliable and more useful for everyday and specialised tasks. The company says the model significantly improves structured responses, document handling, tool use, and multi-step problem-solving compared to GPT-5.1.

GPT-5.2 delivers a major boost in professional-grade reasoning. On multiple evaluations, including GDPval for workplace tasks and ARC-AGI for abstract reasoning, the model outperforms GPT-5.1 by a large margin. OpenAI reports a meaningful drop in hallucinations, with 30% fewer errors on real-world user queries. This helps ChatGPT produce clearer, more reliable answers for research, writing, planning, and technical problem-solving.

One of the biggest upgrades is the long-context window performance. GPT-5.2 can accurately process and reason across hundreds of thousands of tokens, enabling users to upload long documents, research papers, transcripts, and multi-file projects while maintaining accurate references. On OpenAI’s latest MRCRv2 benchmark, GPT-5.2 achieved near-perfect accuracy on deep long-context tasks.

GPT-5.2 Thinking demonstrates significantly better software engineering performance, with state-of-the-art scores on SWE-Bench Pro and Verified. Developers can expect more reliable debugging, refactoring, and implementation of complex features across multiple languages. The model also performs better on front-end work, including UI layouts and even 3D-element designs. This results in fewer manual corrections and faster iteration cycles for engineers.

Vision capabilities receive a substantial lift. GPT-5.2 cuts error rates in half for chart interpretation, GUI screenshots, and technical diagrams. It better understands layout, positioning, and relationships between visual elements, making it more accurate for dashboard analysis, interface documentation, and component identification. For teams working with visual workflows, the model becomes more dependable.

GPT-5.2 introduces stronger agentic behaviour with a new high score of 98.7% on the Tau2-bench Telecom evaluation. For users, this translates to more dependable tool use across long, multi-turn workflows such as customer support resolutions, data retrieval, analysis, and report generation. ChatGPT can now coordinate multi-step tasks with fewer failures and reduced need for user intervention.

GPT-5.2 is now available in ChatGPT across Instant, Thinking, and Pro variants. Users should see clearer responses, stronger structure, and improved handling of complex workloads.