Google has launched Gemini 2.5, its latest advanced AI reasoning model, positioning it to compete directly with rivals like DeepSeek R1 and OpenAI’s latest offerings. According to a blog post released on Tuesday, Gemini 2.5 is currently the highest-rated AI model on LMArena, leading in key areas like reasoning, science, math, and agentic coding.
A New Era of “Thinking Models”
Google describes Gemini 2.5 as a “thinking model”—one that can analyze its own answers recursively before providing a final output, enhancing the model’s accuracy and depth of reasoning. While it leads in many categories, OpenAI’s o3-mini still holds the top spot in LiveCodeBench v5, which tests real-time coding capabilities.
The Gemini 2.5 model is currently available to paid Advanced users, with Google AI Studio providing access on desktop. Some mobile users have reported needing to delete and reinstall the Gemini app to access the updated model.
Key Features and Innovations
One significant advantage Gemini 2.5 holds over its competitors is its high token capacity—the ability to process and understand large and complex data sets. This expanded context window allows for more intricate and nuanced responses across technical and creative tasks.
Users on X (formerly Twitter) have already begun experimenting with Gemini 2.5’s capabilities. In one demonstration, Fei Xia, a Google DeepMind researcher, converted a crude drawing of a three-tier cake into a 3D-printable design. Google also showcased the model’s coding prowess in a video where Gemini 2.5 created a simple endless runner video game in seconds. Another user shared a flight simulation game generated by the model.
Rising Competition in the AI Arena
The release of Gemini 2.5 comes as the AI race heats up globally. The DeepSeek R1 model, released earlier this year by a Chinese research team, delivered an open-source reasoning model that challenged the dominance of American tech giants.
Google is making aggressive investments to maintain its leadership in AI. The company plans to pour $75 billion into AI development in 2025, underscoring its commitment to staying ahead in a market projected to reach $1.8 trillion by 2030, according to Grand View Research.
Beyond Gemini 2.5: Google’s AI Ecosystem
Gemini 2.5 isn’t Google’s only move in the AI space. Earlier this year, it introduced Gemini 2.0 Flash Thinking, a faster reasoning model focused on delivering quick outputs. Last month, Google rolled out Gemini Code Assist, a free AI coding tool offering generous token support for developers.
With AI increasingly integrated across Google’s core products—from Search to Google Docs—the tech giant appears poised to redefine how users interact with technology through generative intelligence.
Google has not yet responded to media requests for further comment.
Leave a Reply