Archive — US — techcrunch
Google launched its deepest AI research agent yet — on the same day OpenAIdropped GPT-5.2
On the same day OpenAI launched GPT-5.2, Google unveiled its advanced research agent, Gemini Deep Research, based on the Gemini 3 Pro model. This tool enhances app integration through a new Interactions API, enabling complex task handling and reducing AI hallucinations. Google also introduced a new benchmark, DeepSearchQA, to test AI capabilities, asserting Gemini's superiority amid rising competition from OpenAI's latest model.
Key points
- Google launched Gemini Deep Research, a reimagined research agent utilizing the Gemini 3 Pro model.
- The agent is designed for complex information synthesis and can be embedded into developers' apps via the new Interactions API.
- It addresses AI hallucinations during complex tasks, designed to ensure accuracy over long decision-making processes.
- Gemini Deep Research aids various applications, from due diligence to drug safety research.
- Google introduced the DeepSearchQA benchmark for testing AI on complex, multi-step information-seeking tasks, which has been open-sourced.
- On the same day, OpenAI launched GPT-5.2, which reportedly outperforms competitors, including Google on various benchmarks.
Why it matters
- The competition between AI models has intensified as both Google and OpenAI strive for leadership in AI development.
- Historically, tech companies often release competitive products around the same time to capture market attention, such as during conferences or major announcements.
- Current IT trends reflect an increasing reliance on AI for complex tasks, reinforcing the narrative of machines gradually assuming decision-making roles.
The launch of Gemini Deep Research amidst OpenAI's GPT-5.2 rollout illustrates the escalating race in AI technology, underscoring the importance of minimizing inaccuracies and maximizing utility as both companies vie for dominance in the future of AI.