OpenAI Unveils GPT-5.2 with Enterprise-Grade Upgrades, Takes Aim at Gemini 3 Pro
The GPT-5.2 family of AI models was made available by OpenAI on Thursday. There are notable performance improvements in the second major update to the fifth generation of the GPT model. According to the business, the AI model in particular provides gains in “economically valuable tasks.” These duties, which include urgent care plans, financial spreadsheets, sales presentations, and more, are basically productivity chores. Additionally, the AI behemoth asserted that GPT-5.2 performed better in multiple benchmark tests than Google’s Gemini 3 Pro. Users will be able to access it through a staggered release.
GPT-5.2 AI Model Released by OpenAI: Features and Availability
The San Francisco-based AI company described the capabilities of the most recent AI model upgrade in a blog post. ChatGPT Go, Plus, Pro, Business, and Enterprise subscribers will be the first to receive the GPT-5.2 AI Models (Instant, Thinking, and Pro). The upgrade will probably be sent to the Edu accounts and the free tier later. Notably, for the next three months, GPT-5.1 will be available to paid users (apart from Go) under legacy models. It will then be retired.
Although Gemini 3 Pro is not mentioned in the main blog post, OpenAI CEO Sam Altman shared a model card of GPT-5.2 to highlight various benchmarks in which it outperforms Google’s frontier model.
Long-context understanding is one of GPT-5.2’s key enhancements. Previous models were able to recall a certain amount of prior text during a session, such as the beginning of a lengthy essay or a business report, but they might become distracted as the content got longer. Larger texts or lengthy discussions can be understood by GPT-5.2 without losing the thread because it is built to retain continuity over lengthier inputs.
Reasoning and job execution are important aspects of the improvement. According to OpenAI, the model excels in multi-step activities like creating a spreadsheet, writing or debugging code, creating a presentation, or following a set of logical steps to address an issue.
In essence, the zero-shot responses have been enhanced, so users won’t need to correct the AI as much through repeated reminders and iterations.
Additionally, GPT-5.2 enhances agentic tool-calling. This implies that the model can more reliably and with better internal checks engage external functions, apps, or tools, like calling a calendar application programming interface (API), connecting with a spreadsheet tool, or initiating code execution.
The model’s vision skills have also been enhanced by the business. Dashboards, product pictures, technical schematics, and visual reports are now easier for it to understand. Additionally, according to OpenAI, the model’s responses in ChatGPT will be more dependable and structured while maintaining a friendlier demeanor during regular use.
The model is being made available as an API for developers in addition to ChatGPT. The base model costs $14 per million output tokens and $1.75 per million input tokens. At $15 for input and $168 for output (per million tokens), the GPT-5.2 Pro is substantially more expensive.
