Google is launching Gemini 3 Flash, making it the default model in the Gemini app

Google today released the fast and cheap Gemini 3 Flash model, based on the Gemini 3 released last month, that looks to steal OpenAI’s voice. The company is also working on making this the default model in the Gemini app and putting AI into the search.

The new Flash model arrives six months after Google announced the Gemini 2.5 Flash model, offering significant improvements. In terms of benchmarks, the Gemini 3 Flash model outperforms its predecessor by a large margin and matches the performance of other frontier models, such as the Gemini 3 Pro and GPT 5.2, on some metrics.

For example, I scored 33.7% without tools on Humanity’s Last Exam, which is designed to test expertise in different fields. In comparison, the Gemini 3 Pro scored 37.5%, the Gemini 2.5 Flash 11%, and the newly released GPT-5.2 34.5%.

In the MMMU-Pro Multimodal Reasoning test, the new model outperformed all competitors with a score of 81.2%.

Consumer proposition

Google is making Gemini 3 Flash the default model in the Gemini app globally, replacing Gemini 2.5 Flash. Users can still choose the professional template from the template chooser for math and programming questions.

The company says the new model is good at identifying multimedia content and gives you an answer based on that. For example, you can upload your own short video about Pickleball and ask for tips; You can try drawing a diagram and have the model guess what you are drawing; Or you can upload an audio recording to get an analysis or create a quiz.

The company also said that the model better understands the intent of users’ queries and can generate more visual answers using elements such as images and tables.

You can also use the new template to create app prototypes in the Gemini app using prompts.

The Gemini 3 Pro is now available to everyone in the US for search and more people in the US can access the Nano Banana Pro photo model in search as well.

Availability of institutions and developers

Google notes that companies like JetBrains, Figma, Cursor, Harvey, and Latitude are already using the Gemini 3 Flash model, available through Vertex AI and Gemini Enterprise.

For developers, the company is making the model available in preview form through an application programming interface (API) and in Antigravity, Google’s new programming tool released last month.

The company said that the Gemini 3 Pro scored 78% in the SWE-certified encoding standard, bested only by GPT-5.2. She added that the model is ideal for video analysis, data mining and visual Q&A, and due to its speed, it is suitable for fast and repetitive workflows.

Model pricing is $0.50 per million input tokens and $3.00 per million output tokens. This is slightly more expensive than the $0.30 per million input codes and $2.50 per million output codes for Gemini Flash 2.5. But Google claims that the new model outperforms the Gemini 2.5 Pro model while being three times faster. For thinking tasks, it uses 30% less code on average than 2.5 Pro. This generally means that you can save the number of tokens for certain tasks.

“We really view Flash as your workhorse model. So, if you look, for example, even at the input and output prices at the top of that table, Flash is just a much cheaper offering from an input and output price perspective. And so it actually allows, for many companies, for batch missions,” Tulsi Doshi, senior director and head of product at Gemini Models, told TechCrunch in a press conference.

Since its release of Gemini 3, Google has processed more than a trillion tokens per day on its API, amid its aggressive rollout and performance war with OpenAI.

Earlier this month, Sam Altman reportedly sent an internal “Code Red” memo to the OpenAI team after ChatGPT traffic declined as Google’s market share among consumers rose. After that, OpenAI released GPT-5.2 and a new image generation model. OpenAI also boasted of its growing enterprise usage and said that the volume of ChatGPT messages has increased 8x since November 2024.

Although Google did not directly address the competition with OpenAI, it said that releasing new models presents a challenge for all companies to be active.

“What’s happening across the industry is that all of these models continue to be great, and challenge each other, and push the boundaries. I think what’s also great is that companies are launching these models,” Doshi said.

“We are also introducing new standards and new ways to evaluate these models. This also encourages us.”