What happened
Google DeepMind is accelerating the development of its next flagship model, Gemini 4, with a target release date set for before the end of 2026. Koray Kavukcuoglu, Google DeepMind's senior vice president, confirmed that the model has entered the phase and is currently undergoing testing on Google’s Antigravity coding platform.
Google DeepMind has officially moved Gemini 4 into the phase, a critical stage where the model undergoes and adjustment based on feedback. According to Koray Kavukcuoglu, the model is currently being tested within Google's internal Antigravity coding environment.
The company is aiming for a release before the end of 2026. This timeline is intended to address the 10-month gap between major releases that has persisted since the launch of Gemini 3 in November 2025.
Google previously faced delays with the Gemini 3.5 Pro model, which missed several projected launch dates throughout the summer of 2026. In response, the company shifted resources toward smaller, faster 'Flash' models to maintain a more consistent cadence of product updates.
The development of Gemini 4 is occurring under a reorganized leadership structure implemented on August 5, 2026. Demis Hassabis has transitioned to chairman of Google DeepMind and chief scientist at Alphabet, while Kavukcuoglu has assumed direct responsibility for model development and the Gemini app teams.
Why it matters
The push for Gemini 4 represents a strategic pivot for Google to shorten its product release cycles. By prioritizing rapid iterations and improvements over the extended delays that impacted previous releases like Gemini 3.5 Pro, Google aims to maintain its competitive standing in the frontier AI market. This shift also reflects a broader organizational focus on deploying functional, agentic AI tools rather than solely pursuing theoretical milestones like artificial general intelligence. The company’s ability to execute this timeline is critical for sustaining its current scale, which includes processing 22 billion tokens per minute across its model APIs.
The shift toward a faster development cycle is a direct response to the competitive pressure of the AI industry. By focusing on rapid iterations, Google hopes to deliver usable AI capabilities to users more frequently, rather than waiting for perfect, large-scale model releases.
Kavukcuoglu emphasized that the company's primary objective is the creation of trustworthy, intelligent agents. This focus on practical utility over abstract goals like AGI suggests a pragmatic approach to product development that prioritizes immediate enterprise and consumer value.
The scale of Google's current operations—supporting over 2.4 million weekly active users on Antigravity and processing 22 billion tokens per minute—underscores the massive infrastructure requirements and the high stakes involved in maintaining a consistent release schedule for its flagship models.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').Which component of an AI application is the machine-learning model itself?
What to watch next
Observers should monitor whether Google meets its year-end target, given the company's history of missed launch windows for previous iterations. Additionally, the performance of Gemini 4 in real-world coding tasks on the Antigravity platform will be a key indicator of the model's capabilities. The effectiveness of the new leadership structure, with Kavukcuoglu overseeing model development and frontier research, remains a significant factor in the project's success.
The primary uncertainty remains whether Google can successfully navigate the phase to meet its year-end deadline, as the release is currently a target rather than a finalized launch.
The impact of the new leadership structure on the speed and quality of model development will be a key area of interest for industry analysts.
Future updates regarding the specific capabilities of Gemini 4, particularly its performance in coding and agentic tasks, will be essential to determine if the model represents a significant leap over the existing Gemini 3.x series.