Google has recently released the preview version of its latest Gemini 2.5 Pro, a model hailed as the "most intelligent" product. It was first announced in March and then upgraded in May. The preview version aims to give enterprises an opportunity to test new applications or replace earlier versions. According to Google's official blog, this new version performs better in responsiveness and creativity, while surpassing other models in coding and reasoning capabilities.
At the I/O Developers Conference in May, Google revealed that the upgrade of Gemini 2.5 Pro significantly enhanced its performance. Demis Hassabis, CEO of Google DeepMind, stated that the I/O version is currently the best coding model offered by the company. However, the Gemini 2.5 Pro Preview 06-05 Thinking released this time surpasses the I/O version and is ready to meet enterprise-level application demands.
The new version of Gemini 2.5 Pro performed excellently in performance tests. Compared to earlier versions, it scored 24 points higher on LMArena and 35 points higher on WebDevArena, ranking at the top of the leaderboard. Google's benchmark tests show that the model outperforms competitors such as OpenAI’s o3, o3-mini, o4-mini, Anthropic’s Claude4 Opus, and xAI’s Grok3 Beta and DeepSeek R1 in coding capabilities.
In its blog, Google mentioned that they have improved the style and structure of Gemini 2.5 Pro based on feedback from previous versions, making its responses more creative and its format more standardized. Although Google continues to improve Gemini 2.5 Pro, this may cause some confusion for many users. The pricing for the new version is $1.25 per million tokens (excluding input caching) and $10 for output.
Since its initial release in March, this model has been widely integrated into many new applications and services, including "Deep Think," which allows Gemini to consider multiple assumptions before responding. After the release of Gemini 2.5 Pro and its two subsequent upgrade versions, Google has regained a foothold in the large language model market, which had previously focused more on the inference models of competitors like DeepSeek and OpenAI.
In just a few hours after the update release, developers began exploring this new version. While many believe the update is indeed faster as promised by Google, further validation of the actual performance improvement of Gemini 2.5 Pro is still needed.
Key points:
🌟 Google has launched a new preview version of Gemini 2.5 Pro, claiming it surpasses major competitors in coding performance.
🚀 The new version scores significantly higher in multiple performance tests, particularly excelling on LMArena and WebDevArena.
💰 Pricing information: $1.25 per million tokens, output price $10; enterprises can use the model on Google AI Studio and Vertex AI platforms.
