Sabrina Ortiz/ZDNET
Google is pushing forward rapidly with updates to Gemini 2.5, the latest addition to the company’s family of AI models, which debuted earlier this year.
On Tuesday, the tech giant announced that Gemini 2.5 Pro and Flash — both of which were originally released as limited public previews — are now generally available and stable. That means they’ve been deemed reliably safe for public use, a win for users. Google also unveiled a new addition to the family — Gemini 2.5 Flash-Lite.
Also: You can now generate images with ChatGPT on WhatsApp – here’s how
Gemini 2.5 Pro and Flash
Gemini 2.5 Pro and Flash are “thinking” models designed to undergo a multi-step reasoning process when responding to user prompts. This process boosts their accuracy and performance, similar to the capabilities found in OpenAI’s o1 and DeepSeek’s R1. Developers can also adjust the models’ “thinking budget,” increasing or decreasing the degree to which they’ll reason through their responses.
(Disclosure: Ziff Davis, ZDNET’s parent company, filed an April 2025 lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
AI models typically undergo rigorous internal safety testing before being deployed to the public. This process has become much more critical in recent years as public and political concerns about the tendency for generative AI models to hallucinate or present false information with the veneer of fact continue to grow. That safety testing is likely why it took Google some time to launch the models into general availability.
Google also recently announced that the limit of daily user prompts for Gemini 2.5 Pro would be doubled from 50 to 100, signaling a growing confidence in the model’s ability to engage in longer conversations.
The models are now accessible through Google AI Studio, Vertex AI, and the Gemini app.
Gemini 2.5 Flash-Lite
Google also introduced on Tuesday a new member of its Gemini family of models called 2.5 Flash-Lite.
Available now in public preview through Google AI Studio and Vertex AI, 2.5 Flash-Lite is designed to be a more cost-effective alternative to its Pro and Flash counterparts. The model has the lowest latency of any of the Gemini models, according to Google, meaning developers won’t have to sacrifice speed for cost. It’s also a reasoning model; however, this capability is off by default to optimize speed. Developers have the option to turn it on by adjusting its thinking budget.
Also: ChatGPT search just got smarter – but can it replace Google for you yet?
The pricing for the Gemini family of models varies depending on the model and the type of content being generated–more info can be found here.
Expanding access
The expansion of Gemini, Google’s flagship generative AI model, into various models of different sizes and capabilities reflects a broader trend underway in Silicon Valley: tech developers are competing to make their tools accessible to as many people as possible.
By releasing models with different capabilities and price points, Google is making it easier for customers to access Gemini regardless of their professional needs or budgets. The company has made Gemini a core part of search with AI Overviews, and the model has also been integrated into Google Meet, Docs, Sheets, and more of the company’s product portfolio.
Also: This new ChatGPT trick lets you download your Canvas work as a PDF in seconds – here’s how
OpenAI has been making a similar effort to maximize the reach of ChatGPT by, for example, integrating the chatbot with WhatsApp.
Want more stories about AI? Sign up for Innovation, our weekly newsletter.
