Google unveiled its new AI reasoning system Gemini 2.5 on Tuesday which improves problem-solving by inserting thinking pauses before creating responses. The integration of more sophisticated reasoning features into Google’s AI models represents a major advancement in technology because it enhances the systems’ precision and contextual comprehension.

The unveiling of Gemini 2.5 Pro Experimental marked the beginning of Google’s announcement of its new AI model family. Experts consider this multimodal AI model the company’s highest-performing system ever developed. Google AI Studio provides access to Gemini 2.5 Pro Experimental starting Tuesday for developers and advanced users. The Gemini app provides access to this feature for subscribers who choose the $20-per-month Gemini Advanced AI plan.

AI Reasoning Models Are Reshaping the Industry

AI reasoning models became a central focus in technology innovation after OpenAI launched its inaugural reasoning model, o1 in September 2024. Leading AI companies, including Anthropic, DeepSeek, xAI, and Google have been actively competing to refine and develop comparable models since the field’s inception. Reasoning-based AI systems use extra computing power and longer processing periods to validate information and improve their problem-solving abilities before delivering an answer.

AI performance in complex fields like mathematics and coding has significantly improved through the implementation of reasoning techniques. Experts throughout the technology industry recognize reasoning models as essential components for creating AI agents that operate autonomously and require minimal human oversight. These models demand higher computational power, which leads to greater operational expenses.

Google’s Efforts to Improve AI Reasoning

Google has previously ventured into the development of reasoning-based artificial intelligence systems before this instance. The company launched an experimental version of Gemini with preliminary “thinking” capabilities in December 2024. Gemini 2.5 stands as Google’s most advanced and competitive model, developed to outdo OpenAI’s esteemed “o” series of models.

According to Google, Gemini 2.5 Pro demonstrates superior performance compared to its previous advanced AI models and multiple top competitor models in various benchmark tests. The company created Gemini 2.5 with the specific purpose of developing visually attractive web interfaces and complex coding-based agent software.

Benchmark Performance: How Gemini 2.5 Pro Compares

The benchmark evaluation of Gemini 2.5 Pro on Aider Polyglot resulted in a code-editing proficiency score of 68.6%. The performance of this system outshines leading AI models from OpenAI and Anthropic along with DeepSeek from China.

Tests with SWE-bench Verified, which evaluates overall software development abilities, found Gemini 2.5 Pro scoring 63.8%. Gemini 2.5 Pro outperforms OpenAI’s o3-mini and DeepSeek’s R1 but doesn’t reach Anthropic’s Claude 3.7 Sonnet which scored 70.3%.

On the Humanity’s Last Exam, which is a multimodal evaluation consisting of thousands of crowdsourced questions from fields like mathematics and humanities, Gemini 2.5 Pro achieved an 18.8% score. The Gemini 2.5 Pro model surpasses the majority of other flagship models available in its category.

Expanding Context Window for Enhanced Processing

The expanded context window serves as one of Gemini 2.5 Pro’s defining features. The system is able to handle as many as 1 million tokens from a single input, which translates to about 750,000 words. The processing capability of Gemini 2.5 Pro exceeds the total word count of the “Lord of the Rings” book series. Google recently revealed its intention to expand processing capacity by making support for 2 million token input lengths available soon.

What’s Next for Gemini 2.5 Pro?

Google has yet to reveal the pricing structure for Gemini 2.5 Pro’s API services. The company plans to provide more information about pricing and commercial availability in the upcoming weeks.

The introduction of Gemini 2.5 demonstrates Google’s strengthened dedication to developing AI reasoning technology. The company’s newest advancements reveal a future where AI models will demonstrate higher sophistication while becoming more intelligent to solve complex problems with enhanced precision as competition grows.