Google is making two major moves this week: launching a new coding-focused AI model and publicly committing to a strategic shift away from chatbots toward AI agents that can autonomously complete tasks. The company's leadership, including CEO Sundar Pichai, is signaling that the future of AI at Google centers on systems that don't just answer questions but actively work alongside humans to accomplish real-world goals.

What Is Google's New Gemini 3.8 Flash Model?

Google's AI research unit plans to ship Gemini 3.8 Flash, internally codenamed "Skimaki," as soon as Wednesday, according to reporting from the Wall Street Journal. The model is specifically designed to improve Google's coding capabilities, an area where the company has lagged behind competitors like Anthropic and OpenAI. Engineers at Google have tested Skimaki against Anthropic's Opus model using Jetski, Google's internal coding tool, and have preferred the new model in those comparisons.

The Flash series is built to be smaller, cheaper, and faster to run than Google's flagship models, which are built from trillions of numerical parameters. This makes it practical for businesses developing autonomous AI systems. Google introduced Gemini 3.7 Flash roughly three weeks before this latest release and saw gains on coding benchmarks including FrontierCode 1.1 and DeepSWE v1.1 compared with its predecessor.

However, a strong showing by Gemini 3.8 Flash would not by itself reestablish Google's position at the frontier of AI development. The company has also struggled with its larger model lineup. Internal candidates for Gemini 3.5 Pro were scrapped because they did not represent a sufficient improvement over the Flash series. Pichai had said in May that a Pro model would arrive "next month," but none has materialized. Gemini 4, Google's next planned flagship, posted encouraging numbers in pretraining evaluations but has yet to finish the posttraining phase.

Why Is Google Pivoting From Chatbots to AI Agents?

Beyond the new model release, Google DeepMind is undergoing a conceptual transformation. The company increasingly views Gemini not as a chatbot or language model, but as an AI agent capable of taking actions on behalf of and alongside humans. This shift reflects Pichai's own vision that agentic AI is the future of search and productivity tools.

"Going from the initial 3.0 launch, my reflection is we learned a lot in terms of understanding what it means to do coding, and not just coding, right? Like what it means to do software engineering, what it means to work with tools or work with the functions that people use every day. Basically turn this whole thing into an agent, from a model to an agent," said Koray Kavukcuoglu, SVP and Chief AI Architect at Google DeepMind.

Koray Kavukcuoglu, SVP and Chief AI Architect at Google DeepMind

Kavukcuoglu explained that coding served as the gateway to understanding software engineering, tool use, and agentic workflows. The goal is not to create a model that answers questions better, but to create something that can take actions and complete tasks.

How Google Is Building AI Agents: Key Strategic Moves

• Reinforcement Learning Investment: Google has devoted more resources to reinforcement learning, a later stage of model training that teaches models to perform skills through trial and error, since the start of the year.
• Leadership and Execution Focus: Demis Hassabis, the unit's co-founder and Nobel Prize winner in chemistry, stepped aside as chief executive of Google DeepMind last month. His successor, Koray Kavukcuoglu, has told employees he wants to increase the pace of execution and has already been directing day-to-day Gemini decisions for at least a year.
• Talent Acquisition in Posttraining: Google brought on Barret Zoph, who previously co-founded Thinking Machines Lab and served as OpenAI's posttraining lead, to fill a vice president of research role covering reinforcement learning and posttraining.

Kavukcuoglu emphasized that what has changed is not the fundamental techniques of AI development, but the environment in which AI operates. Modern AI systems must infer intent, deal with ambiguity, and collaborate with humans in ways that previous generations did not.

The work surrounding Gemini 3.5 taught Google's team about how people actually work with agents. Kavukcuoglu stated he now feels confident that Google has a better understanding of agentic interactions and workflows. This learning process has informed the development of subsequent models, including the upcoming Gemini 3.8 Flash release.

Google's broader product ecosystem reflects this agentic vision. Products like Gmail, Google Sheets, and Maps all help people accomplish tasks rather than simply provide information. Kavukcuoglu explained that AI models are conceptually becoming more than just chatbots; they are becoming agentic systems that reflect how everything at Google is changing, including Search itself.

The timing of these announcements, combined with the leadership transition at Google DeepMind and the hiring of experienced posttraining experts, suggests that Google is making a serious organizational commitment to competing in the emerging agentic AI space. Whether Gemini 3.8 Flash's coding improvements will be enough to shift perceptions of Google's AI capabilities remains to be seen, but the strategic pivot toward agents signals where the company believes the future of AI lies.