Google launches Gemini 3 with big upgrades in reasoning, multimodal performance and agentic tools: All details

After much anticipation, Google has finally announced the next version of its flagship artificial intelligence model, Gemini 3. According to Google’s blog post, the new model improves reasoning, multimodal understanding, and autonomous task execution and is available for use across multiple Google products, including Search, the Gemini app, and developer platforms.

Survey

The Gemini 3 Pro, which was released in preview, is the first model in the new series. Google claims it outperforms its predecessor, Gemini 2.5 Pro, in key industry benchmarks. The company reported high scores on LMArena, Humanity’s Last Exam, GPQA Diamond, and MathArena Apex, as well as improved performance in multimodal assessments such as MMMU-Pro and Video-MMMU. The model also received a high accuracy rating from SimpleQA Verified.

Add As A Trusted Source For Google.

Add as a preferred source on Google

The company has also released a new version, Gemini 3 Deep Think, which is designed for more advanced reasoning tasks. Early testing by the company shows that it performs better on complex problem-solving benchmarks such as Humanity’s Last Exam and ARC-AGI 2. Google stated that this mode will be made available to Google AI Ultra subscribers following additional safety checks.

Gemini 3 integration to Search

For the first time, Google is including its newest model in Search at launch. AI Mode in Search will now use Gemini 3 to create more dynamic layouts, interactive elements, and detailed visual explanations. The model is also accessible via the Gemini API, Google AI Studio, Vertex AI, the Gemini CLI, and a new agent-focused development environment called Google Antigravity.

Antigravity allows developers to use AI agents to plan, write, and validate code within an integrated workspace. The platform includes Gemini 3 Pro, Google’s most recent computer-use model for browser-based actions, as well as an updated image editing model.

According to Google, Gemini 3 also improves long-term planning. The model received the highest score in the most recent Vending-Bench 2 test, which assesses an AI system’s ability to maintain consistent decision-making over extended tasks. These capabilities are also being made available to end users via Gemini Agent within the Gemini app.

Google also announced that additional models in the Gemini 3 lineup will be released in the coming weeks.

Follow Us

Ashish Singh

Ashish Singh is the Chief Copy Editor at Digit. He's been wrangling tech jargon since 2020 (Times Internet, Jagran English '22). When not policing commas, he's likely fueling his gadget habit with coffee, strategising his next virtual race, or plotting a road trip to test the latest in-car tech. He speaks fluent Geek. View Full Profile