Google says Gemini 4 release is coming ‘as soon as possible’

Google has announced that its highly anticipated Gemini 4 AI model is nearing release, according to a statement from DeepMind chief Koray Kavukcuoglu. Speaking at the AI Agenda Live Summit, Kavukcuoglu confirmed that Gemini 4 has entered the post-training phase, which involves refining the model's behavior and implementing safety measures prior to public release. This includes rigorous testing, such as using Gemini 4 to power Antigravity.

Kavukcuoglu expressed excitement about the model and stated that they plan to release an early post-training version as soon as possible, emphasizing their commitment to rapid iterations. While he did not provide a specific timeline, he suggested that the model could be released well before the end of 2026. This announcement comes shortly after major updates from competitors Anthropic and OpenAI, which has raised concerns about Google potentially falling behind.

Kavukcuoglu reassured that the team remains confident and is steadfast in their pursuit of advancing the frontier of AI technology. The company's last major flagship model release was in November 2025 with Gemini 3 Pro, followed by Gemini 3.1 Pro in February 2026. Since then, there have been numerous updates to other models, including Gemini 3.8 Flash.

However, the potential release of Gemini 3.5 Pro remains uncertain, as Kavukcuoglu revealed that Google temporarily shifted its focus to Flash models during this period. When asked about the concept of Artificial General Intelligence (AGI) in relation to other AI model updates, Kavukcuoglu emphasized the importance of building intelligent agents that can be trusted.