Google has launched its newest artificial intelligence model, Gemini 4 Argon. It was released on 13 October 2026 for developers, businesses and everyday users. Google says the model is built for hard, multi-step tasks.
The model is multimodal, which means it can understand many types of content at once. A lawyer could feed it a long contract as a PDF, or a coder could share screenshots of broken code. The model then reasons through the problem step by step.
The model works across coding, finance, legal work, cybersecurity and science. Google says it ranks first on the Vals Index, which tests how well models perform in real economic sectors such as finance and tax work.
Google says it ranks first on the Vals Index, which tests how well models perform in real economic sectors such as finance and tax work.
Big memory and fair pricing
Argon can handle a context window of about one million tokens. That means it can read very long documents in a single go. It accepts text, images, video, audio and PDFs as input, but it only produces text output.
The launch price is 2 dollars per million input tokens and 10 dollars per million output tokens. This low price runs until 30 November 2026, after which the price will double. Cached input tokens get a 95 percent discount.
Who it is for
The model is available through the Gemini API, AI Studio, the Gemini app and AI Mode in Search. It was first announced on 30 September, when only trusted cybersecurity experts got access. Now it is open to everyone.
Google hopes Argon will become the top choice for companies that need AI to handle long, complex jobs. It faces tough competition from new models released by Meta and Anthropic this month.
Comments (0)
Be respectful — comments are moderated.