Google is ready to host Google I/O 2025 – Google I/O 2025 – on 20 May. According to rumors, the incident will unveil several upcoming software updates, including Android 16, Android XR and new Gemini AI features. However, a few days before the incident, the CEO of Google Deepmind, Demis Hasabis could not wait and introduce its “most intelligent model”. Calling it a “pre-i/o gift”, Hasabis said that Gemini 2.5 Pro (preview) now has “large-scale better coding capabilities”.

Google Gemini 2.5 Pro (I/O Edition): New Reform

Google has released a enhanced version of its Gemini 2.5 Pro model, now called “Gemini 2.5 Pro (I/O version)” next to its initial unveiling plan for I/O 2025.

Updated Model-Labeld Gemini-2.5- Pro-Preview-05-06- Promotes major upgradation in its ability to handle programming tasks. According to Google, it provides marked improvement in “code change, code editing, and developing complex agentic workflow”. The version first makes on the 2.5 Pro model, which was launched in late March.

Google described this model as “heavy enthusiasm”, the company chosen it to release it two weeks before expectation. The upgrade is now live within the Gemini app – where canvas workspace stands to achieve the most – and is also available through Mithun API via Google AI Studio and Wartex AI.

Gemini 2.5 Pro AI Model

Gemini 2.5 Pro Gemini 2.5 is the first model in Google. Google says that Each model in Gemini 2.5 lineup – which is still involved – is “made as a model”, which is capable of arguing through its ideas before responding, resulting in performance and correction accuracy. ” With the latest Gemini 2.5 Pro release, that design focus has been carried forward to handle more demand and complex tasks.

The company highlights that the model “LMARENA tops the top of the leaderboard – which measures human preferences – by a significant difference.” It also moves forward in the academic benchmark, taking first place in both Mathematics (AIME 2025) and Science (GPQA Diamond), it is “done without test-time techniques that increase costs, such as majority polls.”

Its standout features have a major upgrade in programming performance, which Google describes as “a large jump more than 2.0”, which has “more improvements”. The model is particularly strong when it comes to creating rich, interactive web apps and agent-powered code tools, in addition to its abilities in “code change and editing”.

In the evaluation focused on autonomous software engineering, Gemini 2.5 Pro, using a equipped agent-based approach, scored an impressive 63.8 percent of 63.8 percent on SWE-Bench verified-a result of which outlines its growing ability in real-world coding challenges.