Google recently unveiled its latest AI reasoning model, Gemini 2.5, on its Google AI Studio developer platform and Gemini app. This model boasts the ability to "think" before responding, marking a significant advancement in AI technology. Google also introduced Gemini 2.5 Pro Experimental, a multimodal reasoning AI model considered the most intelligent to date.
Gemini 2.5 is a new service offered to Gemini Advanced subscribers for $20 per month, allowing users to experience its powerful reasoning capabilities.
Since OpenAI released its first AI reasoning model in September 2024, major tech companies like Anthropic, DeepSeek, Google, and xAI have launched their own reasoning models, vying for supremacy in intelligence. The application of reasoning technology has significantly improved the performance of these models in mathematical and programming tasks, driving further advancements in AI.
Google claims that Gemini 2.5 Pro outperforms its previous AI models and several leading competitors across multiple benchmarks. For instance, in the Aider Polyglot code editing assessment, Gemini 2.5 Pro achieved a high score of 68.6%, surpassing top AI models from OpenAI, Anthropic, and DeepSeek.
In software development capability tests, while scoring 63.8% (lower than Anthropic's Claude 3.7 Sonnet at 70.3%), it still outperformed OpenAI's o3-mini and DeepSeek's R1.
Furthermore, in the "Human's Last Exam," Gemini 2.5 Pro scored 18.8%, again outperforming several competitors. Google states that Gemini 2.5 Pro features a 1 million-token context window, enabling the AI model to process approximately 750,000 words in a single prompt – longer than the entire Lord of the Rings series. Google plans to introduce a 2 million-token context window in the future, further enhancing the model's capabilities.
It's worth noting that this isn't Google's first foray into reasoning models; a "thinking" version of Gemini was released last December. However, Gemini 2.5 is clearly Google's strongest contender yet against OpenAI's o-series models. Google has yet to disclose pricing details for the Gemini 2.5 Pro API.
Official blog: https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/#advanced-coding
Key Highlights:
🌟 Gemini 2.5 and its Pro version demonstrate reasoning capabilities, signifying a new leap in AI technology.
🚀 Gemini 2.5 Pro outperforms multiple competitors in various benchmark tests.
📈 A 2 million-token context window will be introduced, further boosting the AI model's processing power.