GOOGLE WIDENS GEMINI 4 ARGON ACCESS TO PAYING SUBSCRIBERS

Google has widened access to its Gemini 4 Argon AI model, making it available to paying subscribers weeks after a restricted initial release. Google said the rollout would be cautious given the model's capabilities, and Google DeepMind said access was initially limited to cybersecurity defenders. The model is now available in the Gemini app to subscribers on Google's AI Pro and AI Ultra plans, priced from $20 and $99 a month, and globally to enterprises, Android Studio, AI Studio and the Antigravity platform. Free users cannot access it.

Google says Argon is trained to refuse malicious instructions and includes an automated monitor that halts processing if it detects an unintended workflow. Within coding tools, Antigravity and AI Studio restrict Argon to AI Ultra subscribers only, following Google's decision to limit free users of its Flash models to a reduced version called Flash-Lite. Introductory pricing is $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 after 30 November, according to Google. The company says Argon's capabilities are unchanged since launch, citing benchmark results published last month.

Google says it used Argon to cut YouTube comment-system server load by 45% by removing redundant code, and to find bugs undetected since 2018 in a Project Wing drone physics simulator. The company says the model helped optimise Python codebases, patch cybersecurity vulnerabilities, speed up its Gemma AI models, and advance robotics research by helping ground a humanoid device's physics understanding. Google positions Argon against rival models including OpenAI's GPT-6 Astra and Meta's Muse Wonder. The release comes amid wider scrutiny of AI companies, including concerns over data centre construction, unauthorised AI agent actions and AI safety.