Google Gemini 4 Argon released to global developers and enterprise users

Google on Tuesday announced the launch of Gemini 4 Argon for developers and enterprises around the world. This latest artificial intelligence (AI) model is designed to handle coding, research, writing, and other complex tasks across domains. Google claims that Gemini 4 Argon outperforms other cutting-edge AI models in benchmark tests. It topped the Vals index, which evaluates model performance in economically important industries. This new AI model is available through the Gemini application programming interface (API). Google AI subscribers can also access it.

Google Gemini 4 Argon availability and pricing
Google issued a press release on Tuesday confirming that Gemini 4 Argon is now available to developers and enterprises around the world through the Gemini API. Google AI subscribers can access the feature through Google Antigravity, the Gemini app, and AI mode in Search.

Gemini 4 Argon will launch at a price of $2 (roughly Rs. 190) per million input tokens and $10 (roughly Rs. 900) per million output tokens. Cached input tokens are available at a 95% discount.

The promotional period will end on November 30, 2026. Thereafter, Google will charge $4 (approximately 400 rupees) per 1 million input tokens and $20 (approximately 1,800 rupees) per 1 million output tokens.

Google Gemini 4 Argon Capabilities
Google states that the Gemini 4 Argon supercomputer is designed to handle complex tasks across various fields, including coding, finance, law, cybersecurity, and science. It reportedly outperforms other frontier models in benchmark tests and currently holds the top position on the Vals index.

Gemini 4 Argon is also said to excel at solving complex engineering challenges. According to Google, Argon assisted the team behind the “Antigravity” project in increasing the volume of intelligent code submissions by over 35%, while simultaneously reducing the code rollback rate by nearly one-third. Notably, Google highlighted that Argon helped the Chrome team isolate core web-processing functions from a legacy folder containing roughly 6 million lines of code within the 40-million-line Chromium codebase—a task that would typically take months to complete, but was finished in just a few weeks.

Enterprise clients and startups participating in the Google Access Program—including Deloitte, LexisNexis, Ambience Healthcare, and Isomorphic Labs—have been testing the model’s applications in areas such as in-depth research, document synthesis, longitudinal data analysis, and reasoning across large-scale codebases.

Google stated that Argon has optimized the Google Agent Development Kit (ADK), reducing memory session preparation time by 40% and increasing cold-import speeds by 21%. The model has also reportedly optimized the open-source xTB quantum chemistry software package, accelerating molecular structure prediction by 2.3 times. Furthermore, Google noted that Argon delivers a 2.85x speedup in inference for Gemma 4 E2B on A100 GPUs.

Google confirmed that it is expanding the Fairwind program to include hundreds of new partners.