Google launches Gemini 3 Flash: Key differences in latest AI model
Google updates AI with launch of Gemini 3 Flash (photo: Unsplash)
Google has released Gemini 3 Flash, a faster and more affordable version of the Gemini 3 AI model unveiled last month. The company aims to strengthen its position amid competition with OpenAI. The new model is now set as the default in the Gemini app and in Google's AI-powered search mode, according to TechCrunch
What’s known about Gemini 3 Flash
Gemini 3 Flash was released six months after Gemini 2.5 Flash and delivers a noticeable performance boost. Benchmark results show the new model significantly surpasses its predecessor and approaches flagship solutions like Gemini 3 Pro and GPT-5.2 in several metrics.
For example, in the Humanity's Last Exam test, which evaluates expert knowledge across fields, Gemini 3 Flash scored 33.7% without using tools. By comparison:
-
Gemini 3 Pro – 37.5%;
-
Gemini 2.5 Flash – 11%;
-
GPT-5.2 – 34.5%;
In the multimodal reasoning test MMMU-Pro, the new model achieved the highest score among competitors at 81.2%.
Mass rollout for users
Google has made Gemini 3 Flash the default model in the Gemini app worldwide, replacing Gemini 2.5 Flash. Users can still select the Pro version for tasks in mathematics and programming.
The company notes the model handles multimodal content more effectively. Users can:
-
Upload short videos and receive recommendations;
-
Draw sketches for the AI to recognize;
-
Send audio recordings for analysis or quiz generation;
Additionally, Gemini 3 Flash better understands query intent and more frequently produces visual responses, including images and tables. The Gemini app also allows prototyping apps via text prompts.
At the same time, Google launched Gemini 3 Pro in search in the U.S. market, expanding access to the Nano Banana Pro image-generation model.
For businesses and developers
Google reports that companies like JetBrains, Figma, Cursor, Harvey, and Latitude are already using Gemini 3 Flash. The model is accessible through Vertex AI and Gemini Enterprise.
For developers, Gemini 3 Flash is offered as a preview via API and in Antigravity, Google’s new programming tool launched last month.
According to Google, Gemini 3 Flash scores 78% on the SWE-bench test, second only to GPT-5.2. It is suitable for video analysis, data extraction, and visual Q&A, and its high speed makes it convenient for repetitive workflows.
Gemini 3Flash scores 78% in SWE-bench test (photo: TechCrunch)
Price and Performance
The model is priced at:
-
$0.50 per 1M input tokens;
-
$3.00 per 1M output tokens;
This is more expensive than Gemini 2.5 Flash ($0.30 and $2.50, respectively), but Google claims the new model outperforms Gemini 2.5 Pro, runs three times faster, and uses 30% fewer tokens for reasoning tasks. As a result, users can reduce overall computing costs.
''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,'' said Tulsee Doshi, senior director & head of Product for Gemini Models.
Competition with OpenAI
Since the release of Gemini 3, Google processes over 1 trillion tokens per day through its API amid intensifying competition with OpenAI.
Earlier reports said OpenAI CEO Sam Altman sent staff an internal Code Red memo after ChatGPT traffic dropped and Google’s share among users increased.
Following this, OpenAI released GPT-5.2 and a new image generation model, also reporting growth in corporate usage and an eightfold increase in ChatGPT message volume since November 2024.
While Google does not directly comment on competition, the company notes that the AI market is pushing all players to act more aggressively.
