Google's new model, Argon, is only avaliabe to ultra users on Antigravity (arstechnica.com)

Google has launched its advanced AI model, Gemini 4 Argon, though access is currently restricted to paying users on its Antigravity platform. Initially announced with caution due to concerns over potential misuse, Argon's rollout is noteworthy as it highlights Google’s commitment to responsible AI deployment. While the model’s capabilities remain unchanged, Google emphasizes its advanced reasoning and agentic coding features, which are designed to prevent malicious instructions. An automated misalignment monitor is in place to halt errant processes, underscoring the model's sophistication and the inherent risks associated with powerful AI. Argon has shown significant promise in real-world applications within Google, enhancing systems such as YouTube's comment moderation by identifying and removing redundant emoji rendering, resulting in a 45% reduction in server load. Additionally, it has been utilized in Project Wing's proprietary drone physics simulator, uncovering critical bugs from as far back as 2018. The advancements Argon brings to robotics also suggest a measurable enhancement in embodied reasoning, indicating its potential impact across various tech domains. This launch illustrates both the potential and the cautious approach tech giants must take in deploying powerful AI systems.