The tech giant's latest model delivers pro-grade reasoning at lightning speed—and it's now the default for millions of Gemini users worldwide.
Google on Tuesday unveiled Gemini 3 Flash, a new artificial intelligence model that the company says combines frontier-level intelligence with unprecedented speed and efficiency. The release marks the latest salvo in the intensifying AI arms race between Google and OpenAI.
The model is now rolling out globally as the default in the Gemini app, replacing the previous Gemini 2.5 Flash, and is also being deployed in AI Mode in Google Search. The launch comes six months after Google announced the Gemini 2.5 Flash model.
"Today we're introducing Gemini 3 Flash, our latest model with frontier intelligence built for speed at a fraction of the cost," said Logan Kilpatrick, group product manager at Google DeepMind.
Benchmark Performance
The new model has posted impressive results on industry-standard benchmarks. On GPQA Diamond, a PhD-level reasoning test, Gemini 3 Flash scored 90.4 percent. On Humanity's Last Exam, a benchmark designed to test expertise across multiple domains, the model scored 33.7 percent without tool use—putting it in close competition with GPT-5.2's 34.5 percent and not far behind Gemini 3 Pro's 37.5 percent.
On MMMU Pro, a multimodality and reasoning benchmark, Gemini 3 Flash achieved 81.2 percent—comparable to Google's more powerful Gemini 3 Pro model.
Speed and Cost Advantages
Google is positioning the model as a game-changer for developers seeking both performance and affordability. According to the company, Gemini 3 Flash outperforms Gemini 2.5 Pro while being three times faster at a fraction of the cost.
Pricing is set at $0.50 per million input tokens and $3 per million output tokens. While this is slightly higher than Gemini Flash 2.5's $0.30 per million input tokens and $2.50 per million output tokens, Google claims the efficiency gains offset the price increase.
For thinking tasks, the model uses 30 percent fewer tokens on average than Gemini 2.5 Pro, as measured on typical traffic.
Developer and Enterprise Integration
For developers, the model is particularly compelling for coding applications. On SWE-bench Verified, a benchmark evaluating coding agent capabilities, Gemini 3 Flash achieved a score of 78 percent, outperforming not only the 2.5 series but also Gemini 3 Pro.
"We really position Flash as more of your workhorse model," Tulsee Doshi, Senior Director and Head of Product for Gemini Models, told TechCrunch. "It actually allows for, for many companies, bulk tasks."
Companies including JetBrains, Figma, Cursor, Harvey, and Latitude are already using the model. The model is available through Vertex AI and Gemini Enterprise for business customers, and via the Gemini API in Google AI Studio, Gemini CLI, and Google's new agentic development platform Antigravity.
Competition Heats Up
The release comes amid fierce competition with OpenAI. Since launching Gemini 3, Google has been processing over 1 trillion tokens per day on its API.
Earlier this month, OpenAI CEO Sam Altman reportedly sent an internal "code red" memo to his team after ChatGPT's traffic dipped as Google's consumer market share rose. OpenAI has since responded with GPT-5.2 and a new image generation model - ChatGPT Images.
Asked about the rivalry, Doshi noted: "All of these models are continuing to be awesome, challenge each other, push the frontier. And I think what's also awesome is as companies are releasing these models, we're also introducing new benchmarks and new ways of evaluating these models."
Consumer Features
For everyday users, Google says the model excels at understanding multimodal content and generating quick, actionable responses. Users can upload videos for analysis, sketch drawings for the AI to identify in real time, or submit audio recordings to generate quizzes and explanations.
Users can also create app prototypes directly within the Gemini app using simple prompts—a feature Google has dubbed "vibe coding."
Gemini 3 Flash is available now in preview through Google AI Studio, Vertex AI, and the Gemini app. The model represents the third member of Google's Gemini 3 family, joining Gemini 3 Pro and Gemini 3 Deep Think, both released last month.
