Google Launches Gemini 4 Argon AI Model

Google launched Gemini 4 Argon globally on 13 October 2026 after an earlier limited release to testing partners. The artificial intelligence model is designed for complex tasks in software engineering, finance, legal services, cybersecurity and scientific research. Developers and businesses can access it through the Gemini API, AI Studio and Google Antigravity; eligible Google AI Ultra and Pro subscribers can use it through the Gemini app and AI Mode in Search.

Access and pricing

The introductory API price is 2 per million input tokens and 10 per million output tokens. Cached input tokens receive a 95 per cent discount. The introductory prices apply until 30 November 2026, after which standard rates are set to double. A token is a unit of text processed by an AI model; a cached input token is input reused from an earlier request.

Google plans to offer 10,000 scientists six months of free access to the highest usage tier in Google Antigravity. Access is subject to eligibility, capacity and selection. Gemini 4 Argon is the first model in a wider Gemini 4 family.

Benchmarking and applications

Gemini 4 Argon ranked first on the Vals Index, which assesses AI models across economic sectors including finance, coding, legal work and tax work. The model is designed to handle lengthy, multi-step tasks and use multiple tools. Google’s internal testing found that engineers using Argon in Antigravity increased agent-based code submissions by more than 35 per cent and reduced code rollbacks by nearly a third.

The Chrome team used the model for an architectural refactoring project involving hundreds of files, completed in weeks. In scientific research, Argon helped researchers identify enzyme designs that showed catalytic activity in laboratory tests. Cybersecurity teams used it to identify and fix vulnerabilities in sensitive government and healthcare systems.

Safeguards against prompt injection

Prompt injection is an attack in which malicious instructions embedded in content attempt to alter an AI system’s behaviour. Google added safeguards against malicious requests and hidden instructions in webpages, emails and code. On a benchmark involving 15 adaptive attempts, prompt-injection attacks had a reported success rate of 0.7 per cent.

Important Facts for Exams

• Google launched Gemini 4 Argon globally on 13 October 2026.
• The Gemini API provides a way for developers to access Google’s AI models through software applications.
• Enzymes are biological catalysts that accelerate chemical reactions.
• Prompt injection is a type of attack targeting instructions followed by an AI system.

Google CEO Sundar Pichai announced the rollout on X. Google’s reported prompt-injection test involved 15 adaptive attempts.