Google is releasing Gemini 2.5, another major AI update, just months after releasing the Gemini 2.0 Flash Thinking model. Gemini 2.5 Pro Experimental, the first model based on this new technology, is currently accessible to Gemini AI Studio and Advanced users.
Infused with reasoning
Google describes version 2.5 as its “most intelligent” AI model to date in a news release. It is the second of its kind and falls into the same category as Gemini 2.0 Flash Thinking, which is a thinking model. Gemini 2.5 differs from previous models in that it can use contextual intelligence, draw well-reasoned conclusions, and make educated decisions, going beyond the conventional “classification and prediction” method used by non-thinking models.
The Gemini 2.5 model, which was created as an improved version of Gemini 2.0 Flash Thinking, has a better base and more sophisticated post-training. Additionally, Google has shown ambitions to integrate Gemini 2.5’s reasoning capabilities into its non-thinking models.
New benchmark in AI reasoning
Gemini 2.5 Pro Experimental, the first product to use 2.5 version, is offered as a model option within Google’s AI services. Google claims that in the Humanity’s Last Exam test, which is a standard for AI knowledge and reasoning, The2.5 Pro Experimental has fared better than top reasoning models like OpenAI’s o3-mini and Claude 3.5 Sonnet. Furthermore, Gemini 2.5 Pro outperformed rivals like OpenAI’s GPT-4.5 and X’s Grok 3 Beta in math and science benchmarks.
With Gemini 2.5 Pro, Google also highlights improvements in software creation and coding. Coders, editors, and developers can use the concept to create “agentic applications and web apps.” It demonstrated its capabilities by creating a game from a single line of prompt in a demonstration video.
Currently, business and educational plans, as well as Advanced subscribers under One AI Premium, can access Gemini 2.5 Pro. Additionally, it can be accessed through Gemini Studio, and Vertex AI will soon be available as well. Google has said that it will shortly reveal the cost of extended access, even though usage is still restricted.
The release of 2.5 comes after Google opened up the Gemini 2.0 Flash and Deep Research models, as well as Gemini Live screen sharing, which was previously only available to subscribers. It is probable that the AI may soon be widely available due to this tendency.
