A month after launching GPT-5.1, OpenAI is rolling out the GPT-5.2 series (Instant, Thinking, and Pro), which it calls its most advanced solution for professional work. The tool’s creators emphasize it\u2019s moving beyond simple chat to tackle complex, multi-step tasks. The new models are designed to deliver real economic value\u2014from building advanced spreadsheets and writing code to precise analysis of images and long texts.<\/p>
In the GDPval tests, which cover tasks typical of 44 different occupations (representing key sectors of the U.S. economy), the GPT-5.2 Thinking variant matched or outperformed human experts in 70.9% of cases. For comparison, the previous version (GPT-5) hit that mark in only 38.8% of trials.<\/p>
The test tasks focused on producing real deliverables like sales presentations, shift schedules for medical facilities, and production schematics. The judges who reviewed the outputs noted that the documents were strong in both content and presentation, usually needing only minor tweaks. Importantly, the model completed these tasks more than 11 times faster than specialists, at a cost under 1% of the standard market rate for this kind of work (estimates based on historical data).<\/p>
In internal tests covering investment banking analytics tasks, like financial modeling or formatting reports for Fortune 500 companies, the model\u2019s performance jumped from 59.1% (GPT-5.1) to 68.4%.<\/p>
GPT-5.2 Thinking sets a new standard in software engineering. In the rigorous SWE-Bench Pro test, designed to check how well it handles real-world coding problems in four languages, the model scored 55.6%. On SWE-bench Verified, it hit as high as 80%.<\/p>
Early access testers (including teams from companies such as Cognition or JetBrains) say they’re seeing major improvements in front-end development and user interface design, including 3D elements. The model can build a complete interactive application (e.g., a simulation of sea waves or a game) based on a single precise prompt.<\/p>
Its visual skills have gotten better too. It now makes nearly half as many mistakes when interpreting scientific charts and software interfaces. It has a better grasp of spatial relationships, so it can more accurately identify elements in screenshots, technical diagrams, and photos of electronic equipment.<\/p>
One of the key factors for professional use is cutting down on hallucinations. Compared to version 5.1, GPT-5.2 Thinking makes 30% fewer mistakes in its answers (data based on anonymized queries from ChatGPT). It\u2019s still not perfect, but the added accuracy makes it a safer tool for decision-making.<\/p>
There\u2019s been major progress in handling long context, too. In the OpenAI MRCRv2 test, which measures how well a model connects facts scattered across large documents, the model scored nearly 100% accuracy at up to 256,000 tokens. That means GPT-5.2 can comfortably analyze multi-page contracts, research reports and transcripts without losing the thread.<\/p>
New models are rolling out gradually, starting with the paid plans (Plus, Pro, Business, Enterprise). Users get access to three versions:<\/p>
For the API, pricing is $1.75 per million input tokens and $14 per million output tokens (for the base model). Even though the per-unit price is higher than with earlier models, the model\u2019s greater efficiency can ultimately bring down the cost of completing a task, since fewer fixes are needed.<\/p>","protected":false},"excerpt":{"rendered":"
OpenAI is rolling out the GPT-5.2 series, built for professional work. According to OpenAI, the new models handle coding, data analysis and long documents better, while making far fewer mistakes.<\/p>\n","protected":false},"author":230,"featured_media":16688,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"rank_math_lock_modified_date":false,"footnotes":""},"categories":[813],"tags":[707,830,1000,829],"popular":[],"difficulty-level":[],"ppma_author":[884],"class_list":["post-16775","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-news-2","tag-ai-5","tag-chatgpt-2","tag-genai-3","tag-openai-2"],"acf":[],"authors":[{"term_id":884,"user_id":230,"is_guest":0,"slug":"karolina-ceron","display_name":"Karolina Cero\u0144","avatar_url":"https:\/\/haimagazine.com\/wp-content\/uploads\/2025\/07\/PXL_20250419_110132091.MP4-scaled.jpg","first_name":"Karolina","last_name":"Cero\u0144","user_url":"","job_title":"","description":"Wsp\u00f3\u0142tw\u00f3rczyni newslettera AI Flash, studentka psychologii i pasjonatka sztucznej inteligencji. Interesuj\u0119 si\u0119 wp\u0142ywem nowych technologii na cz\u0142owieka, a w wolnych chwilach eksperymentuj\u0119 z generatywn\u0105 grafik\u0105 w Midjourney."}],"_links":{"self":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts\/16775","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/users\/230"}],"replies":[{"embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/comments?post=16775"}],"version-history":[{"count":1,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts\/16775\/revisions"}],"predecessor-version":[{"id":16776,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts\/16775\/revisions\/16776"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/media\/16688"}],"wp:attachment":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/media?parent=16775"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/categories?post=16775"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/tags?post=16775"},{"taxonomy":"popular","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/popular?post=16775"},{"taxonomy":"difficulty-level","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/difficulty-level?post=16775"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/ppma_author?post=16775"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}