Google unveils long-awaited Gemini 4
The model could be a huge catch-up for the company's AI push.
TL;DR
- Google is launching its next-generation AI model, Gemini 4 Argon, to select cybersecurity partners.
- The model is designed for complex, long-running tasks across software engineering, finance, legal work, and cybersecurity.
- Google claims Gemini 4 Argon outperforms OpenAI's GPT-6 Astra on certain coding and knowledge work benchmarks.
- Initial rollout is limited to trusted cybersecurity defenders and government voluntary pre-release programs.
- Access will expand to paying subscribers after additional testing.
- Despite claims of strong performance, some internal testing reportedly found the model lacking, though Google disputes this.
- This release aims to address a perceived gap where Google lagged behind competitors like OpenAI and Anthropic.
- Google's investment in AI safety is highlighted in the context of issues faced by other AI developers.