Google's Gemini 4 Nears Launch as Pressure Mounts to Close AI Gap

Google's upcoming AI model, Gemini 4, is nearing launch as its senior leader Koray Kavukcuoglu confirmed that it has entered 'post-training', a phase where the raw pre-trained weights are refined. This milestone indicates that the core architecture and initial training run for Gemini 4 are locked in place, and researchers can now focus on reliability rather than raw capability.

Kavukcuoglu's comments come as Google is under pressure to release Gemini 4 before the end of 2026. The company has been lagging behind competitors OpenAI and Anthropic, which have shipped new flagship models in recent weeks. OpenAI moved to its sixth-generation Astra model in early September, while Anthropic released two new flagships - Claude Fable 5.1 and Claude Opus 5.5 - within the same three-week window.

Gemini 3.8 Flash, Google's current flagship model, trails behind rivals on independent intelligence benchmarks. According to Valletta Software, Gemini 3.8 Flash scores 41 points, which is significantly lower than OpenAI's GPT-6 Astra and Anthropic's Claude Opus 5.5.

Kavukcuoglu did not confirm a release date or name variants of Gemini 4, such as Gemini 4 Pro or Gemini 4 Flash. The exact features, benchmark scores, context-window length, modality support, and pricing of Gemini 4 remain unknown, as Google has chosen to keep this information under wraps.