Google released a new model of artificial intelligence called Twins 3 Flashwhich is completely free and installed by default in the Gemini mobile app.
Google says that with this version, users will be able to get accurate answers faster.
The company emphasizes that in addition to multimodal capabilities and insight, Gemini 3 Flash has been designed to offer maximum efficiency, surpassing the traditional boundaries between quality, cost and speed.
This model adapts to the complexity level of each task: it can allocate more processing resources for complex cases, but for everyday tasks it uses on average 30% fewer tokens than Gemini 2.5 Pro under normal usage conditions, resulting in better accuracy and performance.
The integration of Gemini 3 Flash into the Gemini mobile app allows users to interact with AI in a more intuitive and dynamic way.
For example, those using the app can upload a personal golf practice video and ask the AI for analysis and suggestions to improve their technique. Gemini 3 Flash’s multimodal capabilities allow it to handle images, video and audio, greatly expanding the range of requests it accepts.
Additionally, users can upload audio, whether it’s a voice recording or a piece of music, to get recommendations, transcriptions, or detailed analysis.
The tool is not limited to text responses, but interprets different formats of information, facilitating access to personalized advice in real time.
Thanks to this integration, the program turns daily or special tasks into simple and efficient processes.
In addition to the Gemini 3 Flash mobile app comes:
- For Gemini API developers in Google AI Studio, Gemini CLI, and the new Google Antigravity agent development platform.
- In AI mode in search.
- For companies on Vertex AI and Gemini Enterprise.
For developers, Gemini 3 Flash integration represents a step forward in iterative development, offering pro-level coding capabilities with low latency.
According to Google, this model is ready to reason and quickly solve tasks in high-frequency workflowsmaking it a key tool for projects that require efficiency and agility.
In the SWE-bench Verified benchmark, which is used to evaluate the performance of encryption agents, Gemini 3 Flash achieved a score of 78%, beating the 2.5 series and Gemini 3 Pro.
This result highlights the optimal balance it offers for the development of coding agents, production-ready systems, and interactive applications where responsiveness is critical.
The company notes that the model also excels in reasoning, tooling, and multimodal processing, making it an ideal choice for developers interested in advanced video analysis, data mining, and complex visual Q&A tasks.
This enables smarter applications.— like helpers in video games or complex A/B testing experiments — that require quick answers and deep reasoning.
Gemini 3 is Google’s most advanced artificial intelligence, integrated directly into Google Search and AI mode.
for example When studying the physics of the three-body problem, you can use interactive simulation and change variables in real time to observe changes in the gravitational interaction.
In the financial sphere, for example when comparing mortgage loans, Gemini 3 in AI mode offers a personalized calculator that compares different options, making it easier to choose the most convenient alternative in the long run.
