Google’s new AI reasoning model Gemini 2.5 debuted on Tuesday, and it boosts problem-solving capabilities by pausing to “think” before producing responses. Google has taken a major leap in AI development by integrating advanced reasoning features into its models to enhance both accuracy and contextual understanding.

Google introduced the Gemini 2.5 Pro Experimental to launch their new family of AI models. The company’s latest multimodal AI system stands as its most intelligent and capable model developed so far. Beginning on Tuesday developers along with expert users will have access to Gemini 2.5 Pro Experimental via Google AI Studio. Gemini Advanced subscribers who pay the company’s $20 monthly AI plan can access it through the Gemini app.

AI Reasoning Models Are Reshaping the Industry

AI reasoning models became a primary industry focus after OpenAI initiated the field with its inaugural reasoning model o1 which launched in September 2024. The development of similar models has become a competitive pursuit among top AI firms including Anthropic, DeepSeek, xAI followed by Google. Reasoning-based AI systems make use of extra computing resources and longer processing durations to verify information multiple times and improve their problem-solving efficiency before delivering their final response.

AI advancements have become evident through the implementation of reasoning techniques which show significant progress in complex fields including mathematics and coding. AI experts predict that reasoning models will become essential for creating AI agents that perform tasks independently with minimal human oversight. The advanced computational needs of these AI models result in higher operating expenses.

Google’s Efforts to Improve AI Reasoning

Google has previously explored the development of reasoning-based artificial intelligence systems. The experimental Gemini version with initial “thinking” features was launched by the company in December 2024. The Gemini 2.5 model stands as Google’s most sophisticated and competitive effort to outperform OpenAI’s esteemed “o” series of models.

Google claims that Gemini 2.5 Pro exceeds the performance levels of both its own previous advanced AI models and many top competitor models, according to multiple benchmark tests. The company purposefully created Gemini 2.5 to produce attractive web applications and handle complex agentic programming tasks.

Benchmark Performance: How Gemini 2.5 Pro Compares

Gemini 2.5 Pro earned a 68.6% score on Aider Polyglot which serves as a benchmark for code-editing skills. The performance metrics of this model exceeded those of leading AI models by OpenAI and Anthropic as well as DeepSeek from China.

Gemini 2.5 Pro received a 63.8% score when evaluated using the SWE-bench Verified benchmark which tests general software development capabilities. Although Gemini 2.5 Pro performs better than OpenAI’s o3-mini and DeepSeek’s R1, it does not match Anthropic’s Claude 3.7 Sonnet which earned a superior score of 70.3%.

Gemini 2.5 Pro obtained an 18.8% score in the Humanity’s Last Exam which combines thousands of crowdsourced questions from fields including mathematics, humanities, and natural sciences. The Gemini 2.5 Pro ranks above most flagship models that compete in this area.

Expanding Context Window for Enhanced Processing

Gemini 2.5 Pro stands out because of its ability to handle a larger context window. The model has the capacity to handle up to 1 million tokens at once which translates to approximately 750,000 words. The Gemini 2.5 Pro’s context window exceeds the total word count of the “Lord of the Rings” book series. Google revealed that it will soon increase its processing capacity to support input lengths of up to 2 million tokens.

What’s Next for Gemini 2.5 Pro?

The pricing details for Gemini 2.5 Pro API have yet to be published by Google. The business plans to provide more information about pricing strategies and when the product will be commercially available in the next few weeks.

Through Gemini 2.5 launch, Google demonstrates its strong dedication to advancing AI reasoning technology. The increasing competition has prompted the company to release advanced AI models, which demonstrate the potential for future systems that deliver more sophisticated intelligence and accuracy when solving complex challenges.