IT之家 news, October 14, Gemini 4 Argon is now available to developers and enterprises worldwide via the AI Studio and Gemini platform APIs, and is also available to Google AI Ultra subscribers on Google Antigravity, and to Google AI Ultra and Pro subscribers on the Gemini app and Search AI Mode.
The model currently has the following limited-time introductory pricing (exchange rate as of October 14, 2026: 1 US dollar (currently about 6.7 RMB) ≈ 6.6921 RMB):
Input: $2 per million Tokens (IT之家 note: currently about 13.4 RMB)
Output: $10 per million Tokens (currently about 66.9 RMB)
Cached input: $0.1 per million Tokens (currently about 0.7 RMB)
After the introductory promotion ends on November 30, 2026, pricing will adjust to:
Input: $4 per million Tokens (currently about 26.8 RMB)
Output: $20 per million Tokens (currently about 133.8 RMB)
Cached input: $0.2 per million Tokens (currently about 1.3 RMB)
Gemini 4 Argon excels at handling complex real-world tasks in coding, finance, law, cybersecurity, science, and other fields, and its depth and precision on multi-step tasks allow it to outperform other frontier models on multiple domain benchmarks. It currently ranks first on the Vals Index, which measures model performance in key economic domains such as finance, coding, law, and tax.
Teams inside Google and at external companies and research institutions have already begun using the model to solve a variety of problems, including refactoring and optimizing large codebases, optimizing business workflows in fields like law and finance, discovering and fixing critical cybersecurity vulnerabilities, and driving new breakthroughs in fields such as quantum computing, nuclear fusion, and the life sciences.
In engineering, the model helped the Chrome team complete an architectural refactoring of tens of millions of lines of code, shortening work that would originally have taken months to just weeks; it also autonomously completed a performance optimization of an open-source project overnight, reducing memory session preparation time by 40% and improving cold import speed by 21%.
In scientific research, Google is providing 10,000 scientists with 6 months of free access to the highest usage tier of Google Antigravity, and the application channel is now open. In cybersecurity, Google has expanded the partners in the Fairwind project to help more cybersecurity defenders use the model to discover and fix vulnerabilities.
On safety, Google has deployed a multi-layered defense system for the model, with significant improvements in jailbreak resistance, prompt injection protection, and agentic alignment monitoring, and the violation rate for harmful requests is nearly half that of the previous-generation Gemini model.
On input and output, the latest Gemini 4 Argon currently supports long decoding outputs of up to 1,048,576 tokens (about 1 million tokens), but input, thinking, and output share the same 1 million token context window.
