Google debuts an updated Gemini 2.5 Pro AI model ahead of I/O
Google Unveils Enhanced Gemini 2.5 Pro Preview (I/O Edition)
On Tuesday, Google announced the launch of the Gemini 2.5 Pro Preview (I/O edition), a refined iteration of its top-tier Gemini 2.5 Pro AI model. The tech giant asserts that this new version outperforms numerous established benchmarks, marking a significant leap in AI capabilities.
The Gemini 2.5 Pro Preview (I/O edition) is accessible through the Gemini API, Google's Vertex AI, and AI Studio platforms. Priced the same as its predecessor, it's also integrated into Google's Gemini chatbot app for both web and mobile users.
Very excited to share the best coding model we’ve ever built! Today we’re launching Gemini 2.5 Pro Preview 'I/O edition' with massively improved coding capabilities. Ranks no.1 on LMArena in Coding and no.1 on the WebDev Arena Leaderboard. It’s especially good at building… pic.twitter.com/9vRaP6RTTo
— Demis Hassabis (@demishassabis) May 6, 2025
The timing of the release aligns with Google's upcoming I/O developer conference, explaining the "I/O edition" tag. At the conference, Google is set to reveal a range of new models, AI-powered tools, and platforms. In the fiercely competitive AI industry, Google is vying for dominance against rivals like OpenAI and xAI, who are also gearing up to launch their highly anticipated models.
Google highlights that the Gemini 2.5 Pro Preview (I/O edition) boasts "significant" enhancements in coding and the development of interactive web applications. The model excels in tasks such as code transformation—modifying code to meet specific objectives—and code editing, according to the company.
Image Credits: Google
In a blog post, Google emphasizes that the Gemini 2.5 Pro Preview (I/O edition) tops the WebDev Arena Leaderboard, a benchmark that assesses the ability to create visually appealing and functional web apps. Additionally, the model demonstrates state-of-the-art performance in video understanding, scoring 84.8% on the VideoMME benchmark.
Google's blog post further states, "For developers already using Gemini 2.5 Pro, this new version will not only improve coding performance but will also address key developer feedback including reducing errors in function calling and improving function calling trigger rates. By default, the model has a real taste for aesthetic web development while maintaining its steerability."
Related article
Google's Latest Gemini AI Model Shows Declining Safety Scores in Testing
Google's internal testing reveals concerning performance dips in its latest AI model's safety protocols compared to previous versions. According to newly published benchmarks, the Gemini 2.5 Flash model demonstrates 4-10% higher rates of guideline vi
Google's Stitch AI Simplifies App Design Process
Google Unveils Stitch AI Design Tool at I/O 2025Google introduced Stitch, its revolutionary AI-powered interface design tool, during the keynote at Google I/O 2025. This innovative solution transforms natural language prompts or reference images into
Google Introduces AI-Powered Tools for Gmail, Docs, and Vids
Google Unveils AI-Powered Workspace Updates at I/O 2025During its annual developer conference, Google has introduced transformative AI enhancements coming to its Workspace suite, fundamentally changing how users interact with Gmail, Docs, and Vids. T
Comments (3)
0/200
![LarryHernández]()
LarryHernández
September 14, 2025 at 6:30:34 PM EDT
Google wirft mal wieder ein neues AI-Modell in den Ring... Ich frag mich, ob Gemini 2.5 wirklich was taugt oder nur Marketing-Gerede ist. Benchmarks können manipuliert werden, oder? 🤨 Hauptsache, es funktioniert besser als dieser peinliche Gemini-Start im Februar!
0
![BrianBaker]()
BrianBaker
August 27, 2025 at 6:36:38 AM EDT
Google's Gemini 2.5 Pro sounds like a beast! 😎 Outperforming benchmarks is cool, but I wonder how it stacks up against real-world tasks. Can't wait to see devs play with it at I/O!
0
![AlbertSmith]()
AlbertSmith
August 23, 2025 at 7:01:21 PM EDT
Google's Gemini 2.5 Pro sounds like a beast! 😎 Outperforming benchmarks is cool, but I’m curious how it stacks up against real-world tasks. Anyone tried it yet?
0
Google Unveils Enhanced Gemini 2.5 Pro Preview (I/O Edition)
On Tuesday, Google announced the launch of the Gemini 2.5 Pro Preview (I/O edition), a refined iteration of its top-tier Gemini 2.5 Pro AI model. The tech giant asserts that this new version outperforms numerous established benchmarks, marking a significant leap in AI capabilities.
The Gemini 2.5 Pro Preview (I/O edition) is accessible through the Gemini API, Google's Vertex AI, and AI Studio platforms. Priced the same as its predecessor, it's also integrated into Google's Gemini chatbot app for both web and mobile users.
Very excited to share the best coding model we’ve ever built! Today we’re launching Gemini 2.5 Pro Preview 'I/O edition' with massively improved coding capabilities. Ranks no.1 on LMArena in Coding and no.1 on the WebDev Arena Leaderboard. It’s especially good at building… pic.twitter.com/9vRaP6RTTo
— Demis Hassabis (@demishassabis) May 6, 2025
The timing of the release aligns with Google's upcoming I/O developer conference, explaining the "I/O edition" tag. At the conference, Google is set to reveal a range of new models, AI-powered tools, and platforms. In the fiercely competitive AI industry, Google is vying for dominance against rivals like OpenAI and xAI, who are also gearing up to launch their highly anticipated models.
Google highlights that the Gemini 2.5 Pro Preview (I/O edition) boasts "significant" enhancements in coding and the development of interactive web applications. The model excels in tasks such as code transformation—modifying code to meet specific objectives—and code editing, according to the company.
Image Credits: Google
In a blog post, Google emphasizes that the Gemini 2.5 Pro Preview (I/O edition) tops the WebDev Arena Leaderboard, a benchmark that assesses the ability to create visually appealing and functional web apps. Additionally, the model demonstrates state-of-the-art performance in video understanding, scoring 84.8% on the VideoMME benchmark.
Google's blog post further states, "For developers already using Gemini 2.5 Pro, this new version will not only improve coding performance but will also address key developer feedback including reducing errors in function calling and improving function calling trigger rates. By default, the model has a real taste for aesthetic web development while maintaining its steerability."
Google's Latest Gemini AI Model Shows Declining Safety Scores in Testing
Google's internal testing reveals concerning performance dips in its latest AI model's safety protocols compared to previous versions. According to newly published benchmarks, the Gemini 2.5 Flash model demonstrates 4-10% higher rates of guideline vi
Google's Stitch AI Simplifies App Design Process
Google Unveils Stitch AI Design Tool at I/O 2025Google introduced Stitch, its revolutionary AI-powered interface design tool, during the keynote at Google I/O 2025. This innovative solution transforms natural language prompts or reference images into
Google Introduces AI-Powered Tools for Gmail, Docs, and Vids
Google Unveils AI-Powered Workspace Updates at I/O 2025During its annual developer conference, Google has introduced transformative AI enhancements coming to its Workspace suite, fundamentally changing how users interact with Gmail, Docs, and Vids. T
September 14, 2025 at 6:30:34 PM EDT
Google wirft mal wieder ein neues AI-Modell in den Ring... Ich frag mich, ob Gemini 2.5 wirklich was taugt oder nur Marketing-Gerede ist. Benchmarks können manipuliert werden, oder? 🤨 Hauptsache, es funktioniert besser als dieser peinliche Gemini-Start im Februar!
0
August 27, 2025 at 6:36:38 AM EDT
Google's Gemini 2.5 Pro sounds like a beast! 😎 Outperforming benchmarks is cool, but I wonder how it stacks up against real-world tasks. Can't wait to see devs play with it at I/O!
0
August 23, 2025 at 7:01:21 PM EDT
Google's Gemini 2.5 Pro sounds like a beast! 😎 Outperforming benchmarks is cool, but I’m curious how it stacks up against real-world tasks. Anyone tried it yet?
0
