Google has just rolled out its Gemini 3 Flash model, aiming to take the lead in the AI space, especially against OpenAI’s offerings. This new model, which builds on the previous Gemini 3, showcases some impressive upgrades in performance, making it a formidable competitor.
In benchmarks, Gemini 3 Flash shows a marked improvement. For example, it scored 33.7% on the “Humanity’s Last Exam,” outpacing its predecessor, Gemini 2.5 Flash, which managed only 11%. Even more interesting is that it can hold its own against top models like Gemini 3 Pro, which scored 37.5%, and GPT-5.2, which earned 34.5%.
The Flash model shines in multimodality and reasoning tests too, achieving an impressive score of 81.2% on the MMMU-Pro benchmark. This makes it particularly effective at handling a variety of content types.
As for user experience, Google plans to make Gemini 3 Flash the default in its apps worldwide. Users will still have the option to access the Pro version for specialized tasks like coding. The new model can analyze different types of media, whether you’re uploading a video for advice, asking for help with a drawing, or even seeking insights from an audio recording.
Gemini 3 Flash is also smarter at understanding user intent, which allows it to offer more detailed and visual responses. You can even create app prototypes directly in the app with this model.
For businesses, companies like JetBrains and Figma are already utilizing Gemini 3 Flash, available through Vertex AI. Developers can access it via a preview API and through Google’s new coding tool, Antigravity.
Financially, the new model comes at a price of $0.50 per million input tokens and $3.00 per million output tokens. This is a slight increase from the previous version, but Google claims it offers better performance and efficiency. In fact, for certain tasks, users may use 30% fewer tokens compared to the Gemini 2.5 Pro model.
Google’s senior director, Tulsee Doshi, emphasized that this model is designed to be cost-effective for businesses, especially for bulk tasks. Since the launch of Gemini 3, Google has been processing over 1 trillion tokens daily.
This competitive landscape is heating up. Just recently, OpenAI’s CEO, Sam Altman, sent a “Code Red” memo due to a decline in ChatGPT traffic, signaling the intense rivalry between these tech giants. Altman highlighted that OpenAI has also introduced new models and experienced significant growth in enterprise usage, with message volumes skyrocketing eightfold since late 2024.
As the AI landscape continues to evolve, both Google and OpenAI are pushing boundaries to enhance their offerings. Both companies are introducing new benchmarks to measure model effectiveness, keeping the competition lively and innovative.
For more details on this competitive landscape, read on sources like TechCrunch.
Source link
ai mode,Google,Google Search
