Google and OpenAI sparked a major AI showdown as Google launched its deepest research agent while OpenAI unveiled GPT-5.2, redefining autonomous AI, advanced reasoning, and enterprise-ready automation.
Published: December 12, 2025 12:55 PM IST
On December 11, 2025, Google and OpenAI both unveiled significant new AI systems that were at least partially research-capable. Both companies were unveiling what they considered to be the largest, most capable AI systems they had ever produced for both research and enterprise use cases.
Gemini Deep Research: Google’s Deepest AI Research Agent
Google has a new large language model called Gemini 3 Pro. The new Gemini model has shown state-of-the-art performance on benchmark tests for factual accuracy and ability.
This new Gemini 3 Pro is being used in a new Gemini Deep Research agent, which is focused on iterative, autonomous research workflows, as opposed to traditional chat-based interaction styles. This new Deep Research agent can be used to make a research plan, search, analyze, and fill gaps in knowledge, and finally return a structured result of the research.
Google is also creating a new Interactions API that can be used by developers to embed Gemini’s Deep Research into apps of their choosing, expanding on Gemini’s vision to be part of every tool.
Google shared a demo of the new Gemini Deep Research agent making a research plan, executing the plan, and finally coming back to summarize and synthesize its findings. The vision of the company with this new agent is to move from just research or question-answering to long-form analysis, multi-step planning, and reasoning with integrated tool use.
The base technology of this system is built from Google’s latest Gemini 3 Pro model, which the company says is currently the most factual and capable AI model in its product suite.
Gemini 3 Pro is supporting new capabilities in coding, math, and natural language understanding, making the new Gemini Deep Research agent a foundation for future AI assistants.
Google also open-sourced the DeepSearchQA benchmark, a new dataset it created to measure research agents.
GPT-5.2: OpenAI’s Most Advanced Generative AI System
OpenAI has a new large language model called GPT-5.2. The new GPT-5.2 model is the most capable large language model in OpenAI’s history.
GPT-5.2 includes new features around reasoning, coding, long context, and vision. This new model has variants with names like GPT-5.2 Instant, GPT-5.2 Thinking, and GPT-5.2 Pro, all the way up to GPT-5.2-Turbo-16k.
OpenAI states the new model is “ready for everything, from everyday tasks to the most complex professional and agentic workflows like data analysis and code generation.”
OpenAI shared benchmarks that it internally uses to test new models for many different tasks in different industries. The benchmarks indicated that the new GPT-5.2 Thinking was at or above human expert levels in a variety of professional tasks. The system can execute workflows like presentations, spreadsheets, and complex planning with the added advantage of massive speed and cost benefits.
The company also provided evidence that the model was hallucinating less, being more reliable on long, multi-step tasks, and being a centerpiece of its productivity and automation vision.
There is a reported “code red” signal that was sent internally at OpenAI to increase the pace of development and catch up to competition like Google’s Gemini products.
In addition to a new model, OpenAI also announced a large investment from Disney, that has also included a large licensing deal.
Also Read:
Disclaimer : This story is auto aggregated by a computer programme and has not been created or edited by DOWNTHENEWS. Publisher: india.com
