Google at present launched its quick and low cost Gemini 3 Flash mannequin, primarily based on the Gemini 3 launched final month, seeking to steal OpenAI’s thunder. The corporate can also be making this the default mannequin within the Gemini app and AI mode in search.
The brand new Flash mannequin arrives six months after Google introduced the Gemini 2.5 Flash model, providing important enhancements. On the benchmark, the Gemini 3 Flash mannequin outperforms its predecessor by a big margin and matches the efficiency of different frontier fashions, like Gemini 3 Professional and GPT 5.2, in some measures.
For example, it scored 33.7% with out software use on Humanity’s Final Examination benchmark, which is designed to check experience throughout totally different domains. As compared, Gemini 3 Professional scored 37.5%, Gemini 2.5 Flash scored 11%, and the newly launched GPT-5.2 scored 34.5%.
On the multimodality and reasoning benchmark MMMU-Professional, the brand new mannequin outscored all opponents with an 81.2% rating.
Client rollout
Google is making Gemini 3 Flash the default mannequin within the Gemini app globally, changing Gemini 2.5 Flash. Customers can nonetheless select the Professional mannequin from the mannequin picker for math and coding questions.
The corporate says the brand new mannequin is sweet at figuring out multimodal content material and providing you with a solution primarily based on that. For example, you possibly can add your pickleball brief video and ask for suggestions; you possibly can strive drawing a sketch and have the mannequin guess what you might be drawing; or you possibly can add an audio recording to get evaluation or generate a quiz.
The corporate additionally mentioned the mannequin higher understands the intent of customers’ queries and might generate extra visible solutions with parts like photos and tables.
Techcrunch occasion
San Francisco
|
October 13-15, 2026
You can even use the brand new mannequin to create app prototypes within the Gemini app utilizing prompts.
The Gemini 3 Professional is now out there to everybody within the U.S. for search and extra folks within the U.S. can entry the Nano Banana Professional picture mannequin in search, as properly.
Enterprise and developer availability
Google famous that firms like JetBrains, Figma, Cursor, Harvey, and Latitude are already utilizing the Gemini 3 Flash mannequin, which is accessible via Vertex AI and Gemini Enterprise.
For builders, the corporate is making the mannequin out there in a preview mannequin via the API and in Antigravity, Google’s new coding software launched final month.
The corporate mentioned the Gemini 3 Professional scores 78% on the SWE-bench verified coding benchmark, solely outperformed by GPT-5.2. It added that the mannequin is good for video evaluation, knowledge extraction, and visible Q&A, and due to its pace, it’s fitted to fast and repeatable workflows.
Mannequin pricing is $0.50 per 1 million enter tokens and $3.00 per 1 million output tokens. That is barely dearer than $0.30 per 1 million enter tokens and $2.50 per 1 million output tokens of Gemini Flash 2.5. However Google claims that the brand new mannequin outperforms the Gemini 2.5 Professional mannequin whereas being 3 times sooner. And, for considering duties, it makes use of 30% fewer tokens on common than 2.5 Professional. Meaning total, you may save on the variety of tokens for sure duties.
“We actually place flash as extra of your workhorse mannequin. So for those who have a look at, for instance, even the enter and output costs on the high of this desk, Flash is only a less expensive providing from an enter and output worth perspective. And so it truly permits for, for a lot of firms, bulk duties,” Tulsee Doshi, senior director & head of Product for Gemini Fashions, advised TechCrunch in a briefing
Because it launched Gemini 3, Google has processed over 1 trillion tokens per day on its API, amid its fierce launch and efficiency battle with OpenAI.
Earlier this month, Sam Altman reportedly despatched an inner “Code Crimson” memo to the OpenAI crew after ChatGPT’s visitors dipped as Google’s market share in customers rose. Publish that, OpenAI has launched GPT-5.2 and a brand new picture technology mannequin. OpenAI additionally boasted about its rising enterprise use and mentioned the ChatGPT messages quantity has grown 8x since November 2024.
Whereas Google didn’t straight tackle the competitors with OpenAI, it mentioned that the discharge of recent fashions is difficult all firms to be lively.
“Nearly what’s taking place throughout the trade is like all of those fashions are persevering with to be superior, problem one another, push the frontier. And I feel what’s additionally superior is as firms are releasing these fashions,” Doshi mentioned.
“We’re additionally introducing new benchmarks and new methods of evaluating these fashions. And in order that’s additionally encouraging us.”
