Join an event that enterprise leaders have been trusted for nearly 20 years. VB Transform brings together people who build real enterprise AI strategies. learn more
Google We have released an updated preview of the “most intelligent” model, the Gemini 2.5 Pro. It was first announced in March, upgraded in May, and will release the same model to general availability in a few weeks.
Enterprises can test build new applications and replace previous versions with updated versions of Gemini 2.5 Pro I/O Edition. Blog post According to Google, its responses are more creative and outperforms other models in coding and reasoning.
At its annual I/O developer conference in May, Google announced that it had updated the Gemini 2.5 Pro to be better than its previous iteration. Google Deepmind CEO Demis Hassabis said the I/O edition is the best coding model ever.
But this new preview, called Gemini 2.5 Pro Preview 06-05 Thought, is even better than the I/O edition. The stable version Google plans to publish is “compatible with enterprise scale features.”
The I/O edition, or Gemini-2.5-Pro-Preview-05-06, was first made available to developers and businesses through Google AI Studio and Vertex AI in May. Gemini 2.5 Pro Preview 06-05 You can access your thoughts through the same platform.
Performance Metrics
This new version of the Gemini 2.5 Pro is even better than the original release.
Google said the new version of the Gemini 2.5 Pro has improved 24 points with Lmarena and 35 points with Webdevarena, and is currently at the top of the leaderboard. The company’s benchmark tests showed that the model outperformed its competitors Openai‘s o3, o3-mini, and o4-mini, Humanity‘s claude 4 opus, grok 3 beta from xai and deepseek R1.
“We’ve also worked on feedback from the previous 2.5 Pro releases, improving style and structure. This could make us more creative with better answers,” Google said in a blog post.
What businesses can expect
Continuous improvements to Google’s Gemini 2.5 Pro can be confusing for many, but Google has previously assembled them in response to community feedback. The new version costs $1.25 per million without caching inputs and $10 at the output price.
When the first version of the Gemini 2.5 Pro was released in March, VentureBeat’s Matt Marshall called it “the smartest model you haven’t used.” Since then, Google has integrated the model into many new applications and services such as “Deep Think” and has looked at multiple hypotheses before Gemini responds.
The release of Gemini 2.5 Pro and its two upgraded versions have revived Google’s location in Google’s large model space as competitors such as Deepseek and Openai distracted the industry’s attention to inference models.
While unveiling the updated Gemini 2.5 Pro for just a few hours, developers are already playing with it. Many have found that they are meeting Google’s promise to make this update faster, but if this latest Gemini 2.5 Pro actually improves performance, the ju judge is still out.
