Google LLC is building on the successful launch of its all-powerful Gemini 3 Pro large language model with the debut of a more lightweight and streamlined version called Gemini 3 Flash.

It’s the successor to Gemini 2.5 Flash, designed for applications that require lower latency and costs, and it’s being rolled out across multiple platforms starting today.

Gemini 3 Flash is built on the same foundation as Gemini 3 Pro, which delivers industry-leading performance in terms of complex reasoning, multimodal and vision-based understanding, agentic artificial intelligence and coding-based tasks. Where it differs is that it’s a more streamlined and efficient version of that model, sacrificing a little performance in order to reduce latency and reduce the costs of AI processing.

Nonetheless, Google said Gemini 3 Flash is still one of its most impressive models for agentic workflows and can power AI agents at less than a quarter of the cost of Gemini 3 Pro, while allowing higher rate limits. More importantly, Gemini 3 Flash surpasses the performance of Gemini 2.5 Flash on numerous benchmarks, Google said, with faster time-to-first-token meaning it’s even more responsive.

Gemini 3 Flash is being made available now to consumers in the Gemini app, where it replaces 2.5 Flash, while developers will be able to access it via platforms including Google AI Studio, Gemini CLI, Vertex AI and Google Antigravity, a new agentic-led application development environment that debuted last month.

Gemini is now quick as a flash

In the Gemini app, Gemini 3 Flash becomes the new default model instead of 2.5 Flash, which means that every single user globally will benefit from its superior performance free of charge.

Google Labs and Google Gemini Vice President Josh Woodward said the improved multimodal reasoning capabilities in Gemini 3 Flash mean that the app can now help people to see, hear and understand any type of information much faster than before. “For example, you can ask Gemini to understand your videos and images and turn that content into a helpful and actionable plan in just a few seconds,” he said.

Users can also ask Gemini to create new applications for them from scratch, even if they don’t have any coding knowledge at all. It truly democratizes application development, Woodward said. Someone can just ask Gemini to help them iterate on an idea, dictate their stream-of-consciousness thoughts and transform their vision into a working prototype right there on their laptop or smartphone, he said.

In addition, Gemini 3 Flash is being made the default model for Google Search’s AI Mode, which responds to user’s searches with detailed, AI-generated summaries to help people discover what they need to know faster. Users can expect faster, more accurate summaries, Woodward said.

He explained that Gemini 3 Flash excels at understanding the nuances of user’s questions, and that this allows it to generate more thoughtful and comprehensive responses based on real-time information rather than stale content. “The result effectively combines research with immediate action: you get an intelligently organized breakdown alongside specific recommendations – at the speed of Search,” he promised.

Split-second responsiveness for AI apps

Developers will also benefit from enhanced performance, with Gemini 3 Flash striking the perfect balance between reasoning and speed for agentic coding tasks and responsive, interactive applications.

It’s available now across all of Google’s major development platforms, meaning developers and their applications will be able to leverage its near real-time multimodal processing capabilities. These span complex video analysis, data extraction and visual questions and answers, and enable Gemini 3 Flash to analyze thousands of documents or video archives and generate the required insights as quick as a flash, said Gemini Senior Director of Product Management Tulsee Doshi.

She explained that Gemini 3 Flash has been designed to eliminate almost completely the lag that’s typically associated with larger models. That ensures split-second responsiveness for customer support agents, in-game assistants and other applications where speed is of the essence, she said.

In terms of its performance, Google has tested Gemini 3 Flash across a number of popular benchmarks. It demonstrated best-in-class scores on PhD-level reasoning and knowledge benchmarks like GPQA Diamond (90.4%) and Humanity’s Last Exam (33.7% without tools), rivaling many much larger frontier models. It also showed leading efficiency, outperforming Gemini 2.5 Pro by generating responses three-times faster at a fraction of the cost, Google said.

“The pace of AI model advancements is fundamentally changing how enterprises can unlock value from their content,” said Yashodha Bhavnani, head of AI at Box Inc. “Gemini 3 Flash shows a relative improvement of 15% in overall accuracy compared to Gemini 2.5 Flash, delivering breakthrough precision on our hardest extraction tasks like handwriting, long-form contracts, and complex financial data. This is a significant jump in performance.”

Developers will also find that Gemini 3 Flash is much more cost-effective than either Gemini 3 Pro or 2.5 Flash, Doshi said. For instance, in the Gemini application programming interface and Vertex AI platforms, the model is priced at just 50 cents per 1 million input tokens and $3 per 1 million output tokens. It also comes with standard context caching, which enables cost reductions of up to 90% in applications with repeated token use above certain thresholds.

Image: SiliconANGLE/Dreamina