Google’s DeepMind chief says Gemini 4 will launch before end‑2026, aiming to close AI gap

Google DeepMind’s new chief, Koray Kavukcuoglu, told The Information the Gemini 4 model is in refinement and will be released "as soon as possible"—well before the end of 2026. The announcement follows a year of delayed flagship releases and a leadership shake‑up at DeepMind.

Key points

• Gemini 4 is in the refinement stage and slated for an early post‑training release.
• Google has not shipped a flagship model since Gemini 3 in November 2025.
• OpenAI’s GPT‑6 and Anthropic’s Mythos have already outperformed Gemini 3.
• A promised Gemini 3.5 Pro update announced at I/O was never delivered.
• DeepMind shifted focus to faster, lower‑power Flash models while Gemini 4 progresses.

What happened

In an interview with The Information, Kavukcuoglu said the company’s intention is to "release an early post‑training output because we see the results and we are excited." He emphasized that the team will continue "fast‑paced iterations" after the early release. The statement came after former DeepMind head Demis Hassabis stepped down in August, having previously signaled "great progress" on Gemini 4.

Google’s internal roadmap had earmarked a Gemini 3.5 Pro update for June following a May I/O announcement by CEO Sundar Pichai. The update never materialized, prompting the division to prioritize the development of less powerful Flash models instead of a mid‑cycle Gemini upgrade.

The backstory

Gemini 3 debuted in November 2025 as Google’s last flagship AI offering. Since then, OpenAI launched GPT‑6 and Anthropic released its Mythos series, both delivering higher performance benchmarks than Gemini 3. Those releases widened the competitive gap and put pressure on Google to accelerate its own roadmap.

DeepMind’s leadership transition, with Hassabis exiting and Kavukcuoglu assuming the helm, coincided with a strategic pivot toward rapid, incremental model releases rather than a single, monolithic upgrade. That shift reflects a broader industry trend of delivering smaller, more frequent improvements to stay relevant in a fast‑moving market.

Competitive implications

Google’s decision to push Gemini 4 early signals a desire to reclaim leadership in large‑scale generative AI. By delivering a refined model before year‑end, Google hopes to narrow the performance gap that OpenAI and Anthropic have opened. The move also re‑establishes a predictable release cadence, which could reassure enterprise customers wary of platform instability.

However, the focus on speed may limit the scope of breakthrough features. Competitors that continue to invest in larger, more capable models may retain an edge in specialized tasks and emerging use cases. Enterprises will weigh the trade‑off between early access to a polished model and the potential benefits of waiting for next‑generation capabilities.

What builders and businesses should do

• Align product roadmaps with the anticipated Gemini 4 launch window to capitalize on early integration opportunities.
• Evaluate current workloads against Gemini 3 benchmarks; identify performance gaps that Gemini 4 could address.
• Monitor OpenAI and Anthropic releases to benchmark feature parity and cost efficiency.
• Prepare fallback strategies in case Gemini 4’s rollout is delayed or its capabilities fall short of expectations.
• Engage with Google’s developer programs to gain early access and influence feature prioritization.

The bottom line

Watch for a formal Gemini 4 release announcement before the close of 2026; its timing will shape enterprise AI adoption strategies and competitive positioning in the coming year.