OpenAI has announced the launch of GPT-5.2, its latest and most advanced artificial intelligence model, positioning it as the company's most capable system for professional and enterprise use.[1][2] The new model series, which is being rolled out to paid ChatGPT subscribers and is available to developers via an API, promises significant advancements in reasoning, coding, long-context understanding, and vision.[3][4] The release comes amid a period of intense competition in the AI sector, with many viewing it as a direct response to the strong performance of rival models like Google's Gemini 3.[1][5][6] OpenAI asserts that GPT-5.2 is designed to unlock greater economic value for users by being demonstrably better at complex, multi-step professional tasks such as creating spreadsheets, building presentations, and writing code.[3][7] The company claims that for the first time, one of its models performs at or above the level of human experts on a range of professional knowledge work tasks.[3][8]

The GPT-5.2 series is comprised of three distinct variants: Instant, Thinking, and Pro, each tailored for different levels of complexity.[1][9][10] The "Instant" model is designed for everyday queries and lightweight tasks, while "Thinking" is intended for more demanding work like in-depth document analysis and complex coding projects.[4][9] "Pro" represents the most powerful version, suitable for high-stakes questions where accuracy is paramount.[4][9] A key focus of this release is the model's enhanced ability to handle long-running and intricate projects from start to finish.[3] OpenAI has highlighted significant improvements in the model's capacity to understand and process long documents, such as contracts and research papers, maintaining near-perfect accuracy across hundreds of thousands of tokens.[3][9] Furthermore, the company reports that GPT-5.2 Thinking produces responses with errors 30% less frequently than its predecessor, GPT-5.1, and has a lower hallucination rate.[2][7] This increased reliability is crucial for professional workflows where factual accuracy is critical.[11][12] The user experience has also been refined, with the models exhibiting a warmer, more conversational tone.[2][10]

OpenAI substantiated its claims of professional-grade performance with a variety of benchmark results. A central piece of evidence is the model's performance on GDPval, a new evaluation created by OpenAI to measure proficiency in well-specified knowledge work tasks across 44 different occupations.[3][13] According to the company, GPT-5.2 Thinking beats or ties top industry professionals in 70.9% of these tasks, a substantial increase from GPT-5.1's 38.8% score.[3][14] OpenAI also notes that the model completed these tasks at over 11 times the speed and at less than 1% of the cost of human experts.[3][15] The new model series also sets new state-of-the-art records in other demanding fields. In mathematics, GPT-5.2 Thinking solved 40.3% of problems on the challenging FrontierMath benchmark and achieved a perfect score on a qualifying exam for the International Mathematical Olympiad.[3][16] For software engineering, it scored a record 55.6% on the SWE-Bench Pro evaluation and is being integrated into developer tools like GitHub Copilot.[3][16][17] The model's abstract reasoning capabilities also showed a massive leap on the ARC-AGI-2 benchmark.[3][5][18]

The launch of GPT-5.2 is a significant strategic move for OpenAI in an increasingly crowded and competitive market. The release followed reports of an internal "code red" at the company, signaling an urgent push to accelerate development and maintain its leadership position against rivals like Google and Anthropic.[4][5][8] The new model appears to be priced competitively with Google's Gemini 3, with API access for the base GPT-5.2 model set at $1.75 per million input tokens and $14 per million output tokens.[1][19][4] The rollout began immediately for developers and paid ChatGPT users on Plus, Pro, Business, and Enterprise plans, with OpenAI providing a three-month transition period where the previous GPT-5.1 model will remain available to subscribers.[1][8][13] While OpenAI has promoted its own benchmark results, early independent analysis and user feedback present a more nuanced picture. Some community-driven leaderboards and tests suggest that while GPT-5.2 is a powerful and clear improvement, it may trail competitors like Anthropic's Claude Opus 4.5 in specific areas, such as blind user preference tests or certain coding benchmarks.[20][21][22]

In conclusion, the arrival of GPT-5.2 marks a clear escalation in the artificial intelligence arms race, with a pronounced focus on practical, professional applications. OpenAI's emphasis on reliability, complex reasoning, and expert-level performance in knowledge work underscores the industry's shift from novelty chatbots to indispensable enterprise tools.[3][23] The model's impressive benchmark scores in coding, mathematics, and abstract reasoning demonstrate significant technical progress.[3][16] However, the competitive landscape is more dynamic than ever, with strong offerings from Google and Anthropic creating a tight race at the frontier of AI development.[5][24] The ultimate success of GPT-5.2 will depend not only on its raw technical capabilities but also on its real-world performance and adoption by the professionals it aims to assist. While some view it as a solid incremental upgrade rather than a revolutionary leap, its enhanced skills in handling complex, long-form tasks position it as a formidable tool for automating and augmenting professional work.[14][25][15]