World
India
On Tuesday, Google unveiled Gemini 2.5, a new family of AI models designed to pause and “think” before answering a question. This latest development marks a significant step in AI technology, as reasoning models become more advanced and widely used.
To introduce this new line of models, Google is launching Gemini 2.5 Pro Experimental, a multimodal AI model that the company claims is its most intelligent yet. This model will be available on Google AI Studio for developers and in the Gemini app for users subscribed to Google’s $20-per-month AI plan, Gemini Advanced.
Moving forward, Google plans to integrate reasoning capabilities into all of its AI models.
AI Reasoning: The Next Big Leap in Technology
Since OpenAI introduced the first AI reasoning model, o1, in September 2024, the tech industry has been in a race to develop even more advanced versions. Companies like Anthropic, DeepSeek, Google, and xAI have all launched AI reasoning models that use additional computing power and time to fact-check and analyze problems before providing answers.
This reasoning ability has helped AI models significantly improve in complex tasks such as math and coding. Many in the industry believe that reasoning models will be a crucial component in developing AI agents—autonomous systems capable of performing tasks with minimal human intervention. However, this added intelligence comes at a cost, as these models require more computing resources and are more expensive to operate.
How Gemini 2.5 Stacks Up Against Competitors
Google has experimented with AI reasoning models in the past, launching a “thinking” version of Gemini in December. However, Gemini 2.5 is the company’s most ambitious attempt yet to outperform OpenAI’s “o” series models.
According to Google, Gemini 2.5 Pro outperforms both its previous AI models and several competing AI models in key benchmarks. The company specifically designed it to excel in building visually appealing web applications and advanced coding tasks.
On a coding evaluation called Aider Polyglot, which measures code editing capabilities, Gemini 2.5 Pro achieved a score of 68.6%, outperforming leading models from OpenAI, Anthropic, and DeepSeek. However, in another software development test, SWE-bench Verified, Gemini 2.5 Pro scored 63.8%. While this was better than OpenAI’s o3-mini and DeepSeek’s R1, it fell short of Anthropic’s Claude 3.7 Sonnet, which scored 70.3%.
Another test, Humanity’s Last Exam, evaluates AI models on thousands of questions spanning mathematics, humanities, and the natural sciences. Google reports that Gemini 2.5 Pro scored 18.8%, surpassing most rival flagship models.
A Massive Context Window for Processing Information
One of the standout features of Gemini 2.5 Pro is its ability to process an extensive amount of information in a single session. The model launches with a 1 million token context window—equivalent to approximately 750,000 words. To put this into perspective, that’s longer than the entire Lord of the Rings book series. Soon, Google plans to double this capacity to support up to 2 million tokens.
Future Plans and Pricing
Google has not yet disclosed API pricing for Gemini 2.5 Pro but has stated that more details will be released in the coming weeks. The company remains focused on refining its AI reasoning technology to maintain a competitive edge in the rapidly evolving AI landscape.
With the introduction of Gemini 2.5, Google is making a bold statement in the AI race, pushing the boundaries of what artificial intelligence can do. As reasoning models continue to evolve, they are expected to play a crucial role in the future of AI-driven applications and autonomous systems.
