Gemini 2.5 Flash, Ironwood Chip, and GenAI Upgrades - Google's New Products at Cloud Next
- Google has announced the release of its new Gemini 2.5 Flash AI model.
- A number of other tools and solutions like the Agent Development Kit for launching AI agents, a new Ironwood chip, and updates to the company's neural networks were also unveiled.
- Samsung and Reddit reported a partnership with Google to integrate the corporation's chatbot into its products.
The neural network will soon be available on the Vertex AI developer platform. It offers ”dynamic and guided” computing, allowing processing times to be adjusted based on the complexity of requests.
”[You can customize] speed, accuracy, and cost balance to meet your specific needs. This flexibility is key to optimizing Flash performance in high-volume, high-value applications,” Google emphasized.
Gemini 2.5 Flash is a ”reasoning” model. Similar to OpenAI's o3-mini and DeepSeek's R1, it spends a little more time responding for additional introspection.
Google claims that 2.5 Flash is ideal for ”high-performance” and real-time applications.
”This workhorse model is optimized specifically for low latency and lower cost,” the corporation claims.
Tool for AI Agents
In addition to the new AI model, Google announced a number of other solutions, including the Agent Development Kit, an open-source toolkit that makes it easier to build AI agents.
Such digital agents could be used for customer service, programming, creating marketing campaigns, and other streamlining of routine activities. The agent will require ”less than 100 lines of code” to develop.
The corporation also unveiled Agent2Agent, a set of standards that will allow AI assistants to work together.
New chip
Additionally, Google said it is expanding its efforts to build its own chips. Its new seventh-generation Ironwood processor is designed to meet the requirements of complex AI models.
The launch is expected in late 2025 for Google Cloud customers. The processor is available in two configurations: a 256-chip cluster and a 9,216-chip cluster.
”Ironwood is our most powerful, performant and energy-efficient TPU. It's purpose-built to scale thinking, inferential AI models,” noted Google Cloud Vice President Amin Wahdat.
Ironwood provides 4,614 TFLOPs of processing power at peak. Each chip has 192 GB of dedicated RAM with bandwidth approaching 7.4 Tbps.
The processor features a specialized SparseCore for data processing. Its architecture is designed to minimize information movement and on-chip latency to save power, Google claims.
Updates to AI models
The company released updates to several content-generating AI models available on the Vertex AI cloud platform.
- Lyria text-to-music model became available for a limited number of users;
- Veo 2 video generator has new editing and visual effects customization capabilities;
- Imagen 3 image generator provides ”significantly” higher performance.
- the corporation also launched a voice cloning feature based on Chirp 3, a model for understanding sound.
Lyria is offered as an alternative to music libraries. It allows users to create songs in different styles and genres;
Chirp 3 is capable of synthesizing speech in about 35 languages. It has been equipped with Instant Custom Voice, which is supposedly capable of cloning a voice based on 10 seconds of audio recording. The mode undergoes a ”thorough vetting” process to confirm ”proper voice authorizations”.
Veo 2 gained the ability to remove background images, logos and objects from existing videos, adjust camera angle and tempo in scenes. ;
Imagen 3 improved the ability to remove objects and repair damaged parts of images. ;
Workspace
Google updates its Workspace suite of cloud-based tools to improve the productivity of AI-powered services.
Added Flows solution to automate multi-step processes like updating spreadsheets and searching for information in documents. It can use Gems chatbots and integrate with Google Drive for data retrieval.
”Just describe what you need in simple language, and Workspace Flows will design and create complex, logically-driven flows,” wrote Yuli Kwon Kim, vice president of product at Google Workspace.
Google Docs will soon have the ability to convert drafts into podcast-style reviews. The upcoming ”Help me refine” feature will offer improved structure and stronger arguments.
Google Sheets will add a ”Help Me Analyze” button that will make recommendations, identify trends, and help create interactive charts.
Google Meet will add a tool called ”Make Notes for Me” to help summarize and reiterate specific video call topics.
Partnerships with Samsung and Reddit
Samsung announced the addition of Google's Gemini artificial intelligence to its Ballie home robot through a partnership with Google Cloud. Users will be able to ask various questions and get answers from the chatbot.
”Through this partnership, Samsung and Google Cloud are redefining the role of artificial intelligence in the home. By combining Gemini's powerful multimodal reasoning with Samsung's AI capabilities in Ballie, we are harnessing the power of open collaboration to usher in a new era of personalized AI companions - ones that move with users, anticipate their needs and interact in a more dynamic and meaningful way than ever before,” commented Yongjae Kim, Executive Director of Samsung's Visual Display Division.
Also announcing the expanded partnership with Google is Reddit. The platform's AI search tool Reddit Answers has received an update thanks to its integration with Gemini. Improvements in relevancy and search speed for users are claimed.
Recall that in March, Google presented a new family of ”reasoning” AI models called Gemini 2.5.
Gemini 2.5 Flash, Ironwood Chip, and GenAI Upgrades - Google's New Products at Cloud Next
- Google has announced the release of its new Gemini 2.5 Flash AI model.
- A number of other tools and solutions like the Agent Development Kit for launching AI agents, a new Ironwood chip, and updates to the company's neural networks were also unveiled.
- Samsung and Reddit reported a partnership with Google to integrate the corporation's chatbot into its products.
The neural network will soon be available on the Vertex AI developer platform. It offers ”dynamic and guided” computing, allowing processing times to be adjusted based on the complexity of requests.
”[You can customize] speed, accuracy, and cost balance to meet your specific needs. This flexibility is key to optimizing Flash performance in high-volume, high-value applications,” Google emphasized.
Gemini 2.5 Flash is a ”reasoning” model. Similar to OpenAI's o3-mini and DeepSeek's R1, it spends a little more time responding for additional introspection.
Google claims that 2.5 Flash is ideal for ”high-performance” and real-time applications.
”This workhorse model is optimized specifically for low latency and lower cost,” the corporation claims.
Tool for AI Agents
In addition to the new AI model, Google announced a number of other solutions, including the Agent Development Kit, an open-source toolkit that makes it easier to build AI agents.
Such digital agents could be used for customer service, programming, creating marketing campaigns, and other streamlining of routine activities. The agent will require ”less than 100 lines of code” to develop.
The corporation also unveiled Agent2Agent, a set of standards that will allow AI assistants to work together.
New chip
Additionally, Google said it is expanding its efforts to build its own chips. Its new seventh-generation Ironwood processor is designed to meet the requirements of complex AI models.
The launch is expected in late 2025 for Google Cloud customers. The processor is available in two configurations: a 256-chip cluster and a 9,216-chip cluster.
”Ironwood is our most powerful, performant and energy-efficient TPU. It's purpose-built to scale thinking, inferential AI models,” noted Google Cloud Vice President Amin Wahdat.
Ironwood provides 4,614 TFLOPs of processing power at peak. Each chip has 192 GB of dedicated RAM with bandwidth approaching 7.4 Tbps.
The processor features a specialized SparseCore for data processing. Its architecture is designed to minimize information movement and on-chip latency to save power, Google claims.
Updates to AI models
The company released updates to several content-generating AI models available on the Vertex AI cloud platform.
- Lyria text-to-music model became available for a limited number of users;
- Veo 2 video generator has new editing and visual effects customization capabilities;
- Imagen 3 image generator provides ”significantly” higher performance.
- the corporation also launched a voice cloning feature based on Chirp 3, a model for understanding sound.
Lyria is offered as an alternative to music libraries. It allows users to create songs in different styles and genres;
Chirp 3 is capable of synthesizing speech in about 35 languages. It has been equipped with Instant Custom Voice, which is supposedly capable of cloning a voice based on 10 seconds of audio recording. The mode undergoes a ”thorough vetting” process to confirm ”proper voice authorizations”.
Veo 2 gained the ability to remove background images, logos and objects from existing videos, adjust camera angle and tempo in scenes. ;
Imagen 3 improved the ability to remove objects and repair damaged parts of images. ;
Workspace
Google updates its Workspace suite of cloud-based tools to improve the productivity of AI-powered services.
Added Flows solution to automate multi-step processes like updating spreadsheets and searching for information in documents. It can use Gems chatbots and integrate with Google Drive for data retrieval.
”Just describe what you need in simple language, and Workspace Flows will design and create complex, logically-driven flows,” wrote Yuli Kwon Kim, vice president of product at Google Workspace.
Google Docs will soon have the ability to convert drafts into podcast-style reviews. The upcoming ”Help me refine” feature will offer improved structure and stronger arguments.
Google Sheets will add a ”Help Me Analyze” button that will make recommendations, identify trends, and help create interactive charts.
Google Meet will add a tool called ”Make Notes for Me” to help summarize and reiterate specific video call topics.
Partnerships with Samsung and Reddit
Samsung announced the addition of Google's Gemini artificial intelligence to its Ballie home robot through a partnership with Google Cloud. Users will be able to ask various questions and get answers from the chatbot.
”Through this partnership, Samsung and Google Cloud are redefining the role of artificial intelligence in the home. By combining Gemini's powerful multimodal reasoning with Samsung's AI capabilities in Ballie, we are harnessing the power of open collaboration to usher in a new era of personalized AI companions - ones that move with users, anticipate their needs and interact in a more dynamic and meaningful way than ever before,” commented Yongjae Kim, Executive Director of Samsung's Visual Display Division.
Also announcing the expanded partnership with Google is Reddit. The platform's AI search tool Reddit Answers has received an update thanks to its integration with Gemini. Improvements in relevancy and search speed for users are claimed.
Recall that in March, Google presented a new family of ”reasoning” AI models called Gemini 2.5.