Google Launches 'Gemini 4 Argon' to Execute Complex AI Tasks
Google unveils Gemini 4 Argon model for complex tasks in programming and law, amid competition with OpenAI
Translated by Qiyas from السياسة — Original source
Google unveiled its advanced artificial intelligence model, Gemini 4 Argon, and made it available to developers and companies worldwide, a move aimed at bolstering its competitiveness against OpenAI and Anthropic, following months of delays and leadership changes within the DeepMind lab.
The Alphabet subsidiary stated that the new model is distinguished by its ability to perform what it describes as “deep work,” executing complex, multi-stage tasks in fields such as programming, finance, law, cybersecurity, and science.
Google set the cost of using the model via its application programming interface (API) at $2 per million input tokens and $10 per million output tokens, with prices set to double starting December 1.
News Services
The model’s launch follows challenges the company faced in developing its AI technologies. Last September, Google announced it was abandoning plans to release Gemini 3.5 Pro, which CEO Sundar Pichai had previously promised to launch in June.
The DeepMind lab underwent extensive restructuring following the resignation of its co-founder and CEO, Demis Hassabis, and the departure of several executives responsible for developing Gemini models.
The company made the new model available to subscription customers through its platforms dedicated to AI app development. Gemini 4 Argon can be accessed by Google AI Pro and Ultra subscribers via the Gemini app and AI-powered search, while access through the Antigraphity coding tool is restricted to Ultra subscribers.
According to Google, teams using the model within the Antigraphity tool saw an increase of over 35% in AI-executed code submissions, alongside a reduction in code rejections by approximately one-third.
Early testers of the model included Deloitte, LexisNexis, Ambience Health Care, and Isomorphic Labs.
In the cybersecurity sector, Google announced the expansion of its Veriwind program to include hundreds of new partners, enabling vetted specialists to obtain a version of the model with built-in safety restrictions removed, whereas access was previously limited to select partners.
The company also unveiled an initiative allowing 10,000 scientists to use the highest tier of Antigraphity services free of charge for six months, as part of efforts to expand the application of AI technologies in scientific and research fields.
