AI News Flash | Google Unveils Gemini 2.5 Pro, Intensifying the AI Race

Hello from the world of AI! 🚀In today's newsletter, we bring you the following top topics:　・Google takes on ChatGPT and Grok, unveiling its latest model "Gemini 2.5 Pro" and adding fresh fuel to the AI race　・Behind the scenes of generative AI: how does AI turn a single prompt into text, images, and audio?　・OpenAI's enterprise market surges — outpacing Anthropic and Google

GOOGLE

Google Takes On ChatGPT, Grok　New Model "Gemini 2.5 Pro" Signals All-In Push in AI Race

Google has released a new version of "Gemini 2.5 Pro," called "Gemini 2.5 Pro Preview (I/O Edition)." It reportedly brings significant accuracy and processing gains for developer features like code conversion, web app building, and video analysis.Beyond more flexible code editing, the new model scored well on the latest industry benchmarks — notably topping the "WebDev Arena Leaderboard," a metric for generating web apps that balance beauty and functionality.The announcement comes just ahead of "Google I/O 2025" and is seen as an effort to showcase Google's technical edge over OpenAI, xAI, and others as generative AI competition accelerates.

Source: Google

Leading Industry Benchmarks With Improved Performance

Google's "Gemini 2.5 Pro (I/O Edition)" shows major gains in software development — better accuracy in code conversion and correction, plus a top ranking on the "WebDev Arena Leaderboard," which measures the ability to generate polished web apps.It also led on video understanding, scoring 84.8% on the "VideoMME" benchmark. Google says the model was refined using user feedback, with practical gains like fewer function-calling errors and a higher execution success rate.

Source: Google

Released to Developers, Signaling an All-In Push Into the AI Race

Google released "Gemini 2.5 Pro Preview (I/O Edition)" to developers via the Gemini API, Vertex AI, and AI Studio, and integrated it into the Gemini chatbot for web and mobile. Pricing stays the same as the previous 2.5 Pro.The timing reflects Google's all-in push into the intensifying AI race. Coming right before Google I/O, it seems aimed at boosting developer-community influence and market share — while signs suggest OpenAI and xAI are also readying new high-performance models.This I/O Edition marks an important step in strengthening developer-facing features and expanding Google's AI ecosystem, expected to further accelerate next-generation AI application development.As the AI race intensifies, Google is putting its technology and platform strengths front and center, aiming to seize the initiative ahead.

Generative AI

Behind the Scenes of Generative AI: Text and Images, All From a Single Line of Command

The tech industry is undergoing a fundamental shift beyond mere feature upgrades — the birth of machine "creativity." At the center of this revolution is a technology known as "generative AI."This technology, producing text, images, and audio from a single prompt, is drawing attention across every industry. So what exactly is generative AI, and why has it become such a hot topic worldwide?

Source: Hyperight

What Is Generative AI? A Technology With the Power to "Imagine," Not Just "Understand"

Until now, AI handled specific tasks like prediction and classification by analyzing data and recognizing patterns. "Generative AI," drawing attention in recent years, takes a different approach — creating new content on its own, based on existing data.Generative AI goes beyond data processing, producing entirely new output almost "as if creating." It applies patterns learned from text and images to generate, from scratch, stories or illustrations that never existed — combining learned knowledge and style into new expression.Crucially, the output isn't a simple imitation of existing data. Generative AI understands the structure and relationships in learned data, and by reconstructing them, creates highly original content — this "creativity" is its greatest strength.A concrete example: generating realistic landscape descriptions or character illustrations from just a few lines of user text. Such capabilities apply not only to creative fields but increasingly to business and marketing.

How Does Generative AI's "Creative Brain" Work?

Behind generative AI's ability to produce natural text and images lies a mechanism that is complex yet systematic.First, generative AI trains on vast, diverse data (text, images, audio, and more), building a "knowledge base" capable of handling a wide range of situations and contexts.At its core is a deep-learning neural network. Modeled after the human brain's neural circuits, it uses billions of parameters to capture complex relationships between data.In the generation process, the model progressively assembles output based on the user's input (prompt) or random initial data. This is how new content — text, images, and more — is generated.Generated results are then evaluated, and the model is retrained and adjusted as needed — the mechanism by which output accuracy and expressiveness continuously improve.

OPEN AI

OpenAI Surges in the Enterprise Market, Pulling Far Ahead of Anthropic and Google

Latest data from fintech company Ramp shows OpenAI has taken a commanding lead over competitors in the U.S. enterprise AI market.As of April 2025, 32.4% of U.S. companies used OpenAI's paid AI services — up sharply from 18.9% in January and 28% in March. Anthropic stayed at just 8%, while Google's AI services plunged from 2.3% to 0.1% over the same period.

Source: Ramp

OpenAI now has more than 2 million enterprise users, double the 1 million as of September 2024. It expects enterprise services to become a major revenue pillar, forecasting sales of $12.7 billion in 2025 and $29.4 billion in 2026.OpenAI is also weighing "AI agents" for programming and research, reportedly with premium plans priced at several thousand dollars a month. It doesn't expect profitability until 2029 or later, and remains in an aggressive investment phase.Ramp's research shows OpenAI is rapidly expanding its enterprise AI market share, establishing leadership in both adoption and revenue.

In Closing

We hope today's content helps with your AI use and information gathering. If useful, please share with interested colleagues or friends. We'll keep curating the topics that matter most. Look forward to our next newsletter. Thank you for your support.