- calendar_today August 8, 2025
Google unveiled Gemini 2.5 on Tuesday as a new collection of AI models focused on improving reasoning abilities. The newest development enables artificial intelligence systems to take a moment for processing, which represents an important progression in AI technology.
Google launches the Gemini 2.5 Pro Experimental as the debut model for its new generation of advanced multimodal reasoning AI systems, which they state represents their smartest creation to date. Google will release this model through Google AI Studio and the Gemini app starting Tuesday, exclusively for users with Gemini Advanced at $20 per month.
AI Reasoning Models: The Next Big Leap in Artificial Intelligence
The tech industry entered a competition to build better reasoning models after OpenAI released its first AI reasoning model, o1, in September 2024. The AI reasoning model space now includes the proprietary models developed by Anthropic, DeepSeek, Google, and xAI. These advanced models utilize additional computing resources to confirm facts and perform in-depth problem analysis prior to delivering their answers.
AI models that utilize reasoning capabilities have shown substantial progress in tackling complex mathematical problems and programming tasks. The majority of specialists in the field agree that these models will become essential components in the creation of AI agents that perform tasks with limited human guidance. The advanced nature of reasoning models leads to higher operating costs.
In December, Google initiated an AI reasoning experiment by releasing a “thinking” version of Gemini. Gemini 2.5 stands as the company’s most sophisticated and committed initiative to rival OpenAI’s “o” series of models.
Performance Benchmarks and Capabilities
Google states that their Gemini 2.5 Pro outperforms earlier AI models as well as some of its primary industry competitors in different performance benchmarks. The company developed this model to specialize in creating visually attractive web applications and autonomous coding programs.
Aider Polyglot evaluation showed Gemini 2.5 Pro scored 68.6% in code editing ability tests. The performance of Gemini 2.5 Pro exceeded the results of top AI models from OpenAI and Anthropic as well as DeepSeek from China.
A separate software development evaluation known as SWE-bench Verified resulted in a 63.8% score for Gemini 2.5 Pro. Gemini 2.5 Pro achieved better results than OpenAI’s o3-mini and DeepSeek’s R1, but its score was still below Anthropic’s Claude 3.7 Sonnet, which reached 70.3%.
Gemini 2.5 Pro achieved an 18.8% score on Humanity’s Last Exam, which included thousands of crowdsourced questions across mathematics, humanities, and natural science,s and outperformed many competing flagship models.
Expanding Context Window and Future Updates
The launch version of Gemini 2.5 Pro includes a 1 million token context window, enabling the AI to handle roughly 750,000 words during one session, which surpasses the length of the “Lord of The Rings” book series. Google revealed its intention to expand the current input capacity by doubling it, which would result in a context window of 2 million tokens.
Google has not revealed the API pricing details for Gemini 2.5 Pro, even though they have presented its advanced features. The company announced that pricing details will be released in the upcoming weeks.
This new Google release demonstrates their commitment to AI reasoning and showcases upcoming developments where AI will become more precise and able to conduct complex problem-solving tasks. Tech companies face increasing competition while they expand AI capabilities, which positions reasoning models as essential elements for future intelligent systems.
