To introduce its new family of AI models, Google is launching Gemini 2.5 Pro Experimental—its most advanced multimodal reasoning AI yet. Available starting Tuesday, the model will be accessible through Google AI Studio and the Gemini app for subscribers of the $20-a-month Gemini Advanced plan.

Figure 1. Next-Generation AI Reasoning Models Unveiled by Google.

Looking ahead, Google plans to integrate reasoning capabilities into all its future AI models. Figure 1 shows Next-Generation AI Reasoning Models Unveiled by Google.

Since OpenAI introduced the first AI reasoning model, o1, in September 2024, the tech industry has been in a race to match or surpass its capabilities. Today, major players like Anthropic, DeepSeek, Google, and xAI all have AI reasoning models, which leverage additional computing power and time to enhance fact-checking and problem-solving before generating responses.

These reasoning techniques have significantly improved AI performance in math and coding tasks. Many experts believe reasoning models will be essential for AI agents—autonomous systems capable of operating with minimal human input. However, the increased complexity also makes these models more expensive to run.

Google has previously experimented with AI reasoning, including a "thinking" version of Gemini released in December. Now, with Gemini 2.5, the company is making its most ambitious push to rival OpenAI's "o" series.

Google claims Gemini 2.5 Pro outperforms both its earlier frontier AI models and several top competitors in key benchmarks. The model is designed to excel at visually compelling web app creation and agentic coding applications.

On the Aider Polyglot evaluation, which measures code editing performance, Gemini 2.5 Pro scored 68.6%, surpassing models from OpenAI, Anthropic, and DeepSeek. However, in the SWE-bench Verified test for software development skills, it achieved 63.8%, outperforming OpenAI’s o3-mini and DeepSeek’s R1 but falling short of Anthropic’s Claude 3.7 Sonnet, which scored 70.3%.

On Humanity’s Last Exam—a rigorous multimodal test featuring thousands of crowdsourced questions across mathematics, humanities, and natural sciences—Google reports that Gemini 2.5 Pro scores 18.8%, outperforming most rival flagship models.

At launch, Gemini 2.5 Pro comes with a 1 million-token context window, allowing it to process approximately 750,000 words in a single input—longer than The Lord of the Rings trilogy. Google also plans to expand this capability soon, doubling the input length to 2 million tokens.

Google has yet to disclose API pricing for Gemini 2.5 Pro but says more details will be shared in the coming weeks.

The Rise of AI Reasoning Models

AI reasoning models represent the next frontier in artificial intelligence, enabling systems to analyze, fact-check, and reason through problems before delivering responses. Since OpenAI launched the first AI reasoning model, o1, in September 2024, competitors like Anthropic, DeepSeek, Google, and xAI have entered the race, each developing models with advanced problem-solving capabilities.

Google’s Gemini 2.5 Pro – The Next Leap in AI

To spearhead its new family of models, Google is introducing Gemini 2.5 Pro Experimental, its most sophisticated reasoning AI yet. This multimodal model is designed to excel in complex tasks, including web app development and agentic coding. Available through Google AI Studio and the Gemini app (for $20/month Gemini Advanced subscribers), Gemini 2.5 marks a significant step in AI evolution.

Unprecedented Context Window & Future Expansion

Gemini 2.5 Pro ships with a groundbreaking 1 million-token context window, allowing it to process around 750,000 words—more than the entire Lord of the Rings trilogy. Google plans to double this capacity soon, extending input length to 2 million tokens, making Gemini 2.5 one of the most memory-extensive AI models ever built.

The Road Ahead – AI Agents & Pricing Uncertainty

Reasoning models like Gemini 2.5 Pro are paving the way for AI agents—autonomous systems capable of performing tasks with minimal human oversight. While Google has yet to announce API pricing, it plans to share more details soon. With reasoning capabilities becoming a core feature of future AI models, the competition to build the most advanced, cost-effective AI continues to heat up.

Source: TC

Cite this article:

Priyadharshini S (2025), Google Introduces a New Generation of AI Reasoning Models, AnaTechMaz, pp. 582