Google’s most powerful AI seizes business opportunities for enterprises! The actual measured success rate increases by 20% and information security loopholes can also be discovered

Google launched Gemini 4 Argon, a new generation AI model, and announced that it will be introduced into its enterprise services to help enterprises handle complex tasks that would otherwise take hours or even days to complete. Currently, companies such as ServiceNow, Harvey, Figma and HubX have participated in early testing. Among them, ServiceNow stated that the success rate of Argon in specific enterprise tasks is about 20% higher than that of Gemini 3.8 Flash, especially in the fields of human resources and customer service.

Google Cloud stated that the need for enterprises to introduce AI is shifting from answering questions and writing content to performing complex and multi-step tasks. Gemini 4 Argon can read long documents containing text, tables, charts and pictures, integrate different data sources, conduct long-term trend analysis, and produce analysis results that can be used by enterprises. The application scope covers finance, law, software development and information security protection.

In the financial field, Argon can assist with multi-step accounting reconciliations, spreadsheet formulas and variance analysis, and emphasize the accuracy of calculations, sources cited and formulas. In the legal section, legal technology company Harvey stated that Argon's performance in legal work related to corporate mergers and acquisitions, data privacy, cybersecurity and venture capital has been significantly improved compared to Gemini 3.1 Pro, and it also maintains a certain degree of accuracy in long-form document analysis. Through AI-assisted analysis of precedents, regulatory documents and legal citations, legal professionals can devote more time to case judgment and research.

In the field of human resources and customer service, enterprise software company ServiceNow stated that Argon's task success rate in the early testing of the AI ​​performance evaluation tool NowAI-Bench was about 20% higher than that of Gemini 3.8 Flash, with improvements in human resources and customer service being particularly prominent. ServiceNow introduces Google's new model capabilities into its AI platform to help companies convert some service desk work to be performed autonomously by AI.

Google also hopes to use Argon to help companies accelerate product development. The design platform Figma said that Argon has been used in Figma Make to help users convert complex, multi-step creative ideas into product prototypes. Google's internal open source agent development kit ADK team also uses Argon to improve system performance, shortening the memory preparation time of long-term work phases by 40%, and increasing cold start import speed by 21%.

Information security is another application focus. Google pointed out that Gemini 4 Argon is its model with the strongest ability to resist "indirect prompt injection" attacks so far, which can prevent attackers from inducing the AI ​​to deviate from its original task through malicious instructions or the use of situational information.

In addition, in the information security vulnerability repair benchmark CWE-Bench v1, Argon tied for first place with a score of 68%; in the indirect prompt word injection attack test by Gray Swan, an artificial intelligence security company, the attack success rate was only 0.7% in the face of 15 adaptive attack attempts. In addition, the information security company Wiz has also introduced Argon into the "Scan for Good" program to help protect critical infrastructure for free.