Google Gemini 4 Argon Released to Paid Users with Tiered Features

Google is pitching Gemini 4 Argon as a single, general-purpose AI model to manage the full robotics pipeline, controlling robots, building virtual practice worlds, and grading training videos. For instance, when controlling a Boston Dynamics Spot robot to toss soda cans, Argon first builds a virtual copy of the bin and uses simulation practice before real-world execution, with two extra AI agents verifying positioning. From a single photo of party props, Argon generated over 4,500 training episodes, boosting the real robot's success rate from 56% to 75% and reducing execution time from 35 to 24 seconds. Argon can also generate simulations from teleoperated recordings, first-person videos, or even ballpoint pen sketches, aiming toward 'physical AGI'.

Google made Gemini 4 Argon generally available for coding, knowledge work, enterprise workflows, and cyber defense. Access is available via the paid Gemini API, Gemini platform, Gemini app (with Google AI Pro and Ultra in AI Mode in Search), and Ultra in Antigravity. Introductory API rates are $2/$10 per million input/output tokens, doubling to $4/$20 after November 30. Internally, Google claims Argon helped separate core webpage handling from a six-million-line legacy Chromium folder, completing months of work in weeks, and reported 35% more agentic code submissions with nearly one-third reduction in rollbacks. Argon is also available in Cursor, Devin Desktop/CLI, and GitHub Copilot.

llm-gemini 0.35 now supports the Gemini 4 Argon model, available via the Gemini API and for $100/month subscribers. Argon launched at $2/million input and $10/million output, priced lower than Claude Opus 5.5 and GPT-6 Astra, but is set to double in price on December 1st. Artificial Analysis benchmark data shows Argon costs $2,407 to run, making it significantly cheaper than OpenAI, Anthropic, and Meta's latest models.

Google's new AI model, Gemini 4 Argon, is now globally available, offering enhanced capabilities for developers, businesses, and consumers. Replacing the Gemini Pro series, Argon is accessible through various platforms and boasts improved performance in coding and 3D model generation. However, it trails competitors in some rankings and is priced at $2 per million input tokens.

Google DeepMind has released Gemini 4 Argon, a new frontier AI model now available globally via API, AI Studio, and the Gemini platform. Argon excels at multi‑step tasks in coding, finance, legal, cybersecurity, and science, topping the Vals Index and outperforming other frontier models. It is priced at $2 per million input tokens and $10 per million output tokens, with a 95% discount on cached inputs. Early adopters at Google and enterprises are using Argon to refactor massive codebases, optimize business processes, patch security flaws, and accelerate research in quantum computing and life sciences.

Google is beginning a wide rollout of Gemini 4 Argon in the Gemini app and Antigravity, but access is currently restricted to paying users, and even lower-tier subscribers will not have full access. Google maintains that Argon's capabilities, evidenced by late last month's benchmarks, include advanced reasoning and agentic coding, and it is trained to refuse malicious instructions with an automated misalignment monitor. Argon has been used internally by thousands of Googlers, helping to make YouTube's comment system more efficient by identifying redundant emoji rendering that dropped server load by 45 percent, and finding critical bugs in Project Wing's drone physics simulator dating back to 2018.