Google unveils Gemini 2.5: Its most powerful 'thinking' AI yet – Firstpost
Google has announced the launch of Gemini 2.5, a new family of artificial intelligence (AI) models designed to enhance reasoning capabilities to approach human-level cognition. At the core of this release is the Gemini 2.5 Pro Experimental, touted as the most advanced model by the company.
The Gemini 2.5 Pro Experimental is now accessible through Google AI Studio, the company's developer platform, and the Gemini app for Gemini Advanced users, which is a premium service priced at $20 per month. You can find more information about this on TechCrunch.
Advancements in AI Reasoning
Google's new model enters the competitive arena of AI technologies striving to not only generate text or images but also to incorporate reasoning and fact-checking abilities before responding. This trend was initiated by OpenAI with the launch of its "o1 model" in September 2024, which introduced AI reasoning to a broader audience. Various companies such as Anthropic, DeepSeek, Google, and xAI have since introduced their reasoning-based systems, enhancing problem-solving capabilities, especially in mathematics and coding, with improved accuracy.
Google has declared that all its future AI models will integrate these reasoning techniques as a default practice. While previous versions of Gemini had experimented with such features, the latest release marks a significant leap forward in advancing this technology.
Performance and Benchmarks
Initial benchmarks indicate promising results for Google's Gemini 2.5 Pro. In a coding evaluation focused on code editing tasks called Aider Polyglot, the model scored 68.6%, outperforming top models from other key players. However, in software engineering capabilities evaluation on SWE-bench Verified, it fell short of competitor scores.
Gemini 2.5 Pro, surpassing its predecessor Gemini 2.0, showcased strong performance on a comprehensive test named "Humanity's Last Exam," outscoring many competing models. An impressive feature of this model is its context window, enabling processing up to 1 million tokens in a single operation - equivalent to processing a vast amount of text. Google is planning to double this capacity in the near future.
While Gemini 2.5 Pro has been released, the pricing for API access has not been disclosed by Google yet, with more information expected to be revealed in the upcoming weeks.
