Google has officially launched Gemini 3 Flash, a faster and more cost-effective AI model, positioning it as the new default within the Gemini app and for AI-powered search. Building on last month's Gemini 3 release, this update aims to challenge competitors like OpenAI by offering significant performance improvements over its predecessor, Gemini 2.5 Flash, and matching leading models like Gemini 3 Pro and GPT 5.2 in certain benchmarks.
In benchmark testing, Gemini 3 Flash demonstrated strong capabilities. It achieved a 33.7% score without tool use on "Humanity's Last Exam," a benchmark designed to assess expertise across diverse domains. This compares favorably to Gemini 3 Pro's 37.5% and the recently released GPT-5.2's 34.5%, significantly surpassing Gemini 2.5 Flash's 11%. Notably, on the MMMU-Pro multimodality and reasoning benchmark, Gemini 3 Flash led all competitors with an impressive 81.2%.
Consumer Availability & Features
Google is rolling out Gemini 3 Flash as the new default model in the Gemini app worldwide, replacing Gemini 2.5 Flash. Users will still have the option to select the more powerful Pro model for complex tasks like advanced math and coding questions.
The company highlights Gemini 3 Flash's enhanced ability to process and respond to multimodal content. Examples include analyzing a pickleball video for tips, identifying objects from a user's sketch, or generating quizzes from audio recordings. The model also promises improved understanding of user intent, leading to richer, more visual responses incorporating elements like images and tables. Additionally, users can leverage Gemini 3 Flash within the Gemini app to create app prototypes using simple prompts.
Separately, Gemini 3 Pro is now broadly accessible for search in the U.S., alongside wider access to the Nano Banana Pro image model within search.
Enterprise and Developer Access
Gemini 3 Flash is already being utilized by major companies such as JetBrains, Figma, Cursor, Harvey, and Latitude, accessible via Google's Vertex AI and Gemini Enterprise platforms.
For developers, the model is available in a preview through the API and integrated into Antigravity, Google's new coding tool released last month. Google also noted that Gemini 3 Pro achieved a 78% score on the SWE-bench verified coding benchmark, trailing only GPT-5.2. This makes the model particularly well-suited for tasks requiring rapid and repeatable workflows, such as video analysis, data extraction, and visual Q&A, thanks to its speed and efficiency.
Pricing and Efficiency
Gemini 3 Flash is priced at $0.50 per 1 million input tokens and $3.00 per 1 million output tokens. While this is a slight increase compared to Gemini 2.5 Flash's $0.30 per 1 million input and $2.50 per 1 million output tokens, Google asserts that the new model delivers superior performance to Gemini 2.5 Pro while operating three times faster. Furthermore, for "thinking tasks," it consumes 30% fewer tokens on average than 2.5 Pro, potentially leading to overall cost savings for specific applications.
“We really position flash as more of your workhorse model. So if you look at, for example, even the input and output prices at the top of this table, Flash is just a much cheaper offering from an input and output price perspective. And so it actually allows for, for many companies, bulk tasks,” Tulsee Doshi, Senior Director & Head of Product for Gemini Models, told TechCrunch in a briefing.
The AI Competitive Landscape
Google's release of Gemini 3 Flash comes amidst an intensifying "AI war" with OpenAI. Since the launch of Gemini 3, Google's API has processed over 1 trillion tokens daily. This fierce competition was underscored by reports earlier this month of an internal “Code Red” memo from OpenAI CEO Sam Altman, following a reported dip in ChatGPT traffic and a rise in Google's consumer market share. In response, OpenAI has since released GPT-5.2 and a new image generation model, while also highlighting an eightfold increase in ChatGPT message volume for enterprise users since November 2024.
While Google did not directly comment on the rivalry with OpenAI, Tulsee Doshi acknowledged the dynamic nature of the industry:
“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. And so that’s also encouraging us.”
