|
Google has launched Gemini 4 Argon globally for developers, businesses and eligible subscribers, positioning the new artificial intelligence model for complex tasks across software engineering, finance, legal services, cybersecurity and scientific research.
The model is available through the Gemini API, AI Studio and Google Antigravity for eligible users, while Google AI Ultra and Pro subscribers can access it through the Gemini app and AI Mode in Search. According to a Google blog post, the model has been launched at an introductory price of USD 2 per million input tokens and USD 10 per million output tokens, with a 95 per cent discount on cached input tokens. Google said Argon ranked first on the Vals Index, which evaluates AI model performance across economic sectors including finance, coding, legal and tax work. The company said the model is designed to handle lengthy, multi-step tasks and combine reasoning with the use of multiple tools. Google further reported that engineers using Argon in Antigravity increased agent-based code submissions by more than 35 per cent while reducing code rollbacks by nearly a third. The model also helped the Chrome team complete an architectural refactoring project across hundreds of files in weeks, a task the company said would otherwise have taken months. Google CEO Sundar Pichai highlighted the models coding and tool-use capabilities in a post on X. It has a much stronger grasp of coding and multi-step tool use, he said. He further described how Argon created an exploded-view animation, lighting, camera movements and audio from a raw Pixel computer-aided design file. Pichai said the model ran 1,000 iterative rounds to check its renders and refine the code. The company also cited applications in scientific research and cybersecurity. Google said Argon helped researchers identify new enzyme designs that subsequently showed catalytic activity in laboratory tests, while cybersecurity teams used it to identify and fix vulnerabilities affecting sensitive government and healthcare systems. Google further said it had introduced additional safeguards for the model, including measures against malicious requests and hidden instructions in webpages, emails and code. The company reported that such prompt-injection attacks had a success rate of 0.7 per cent on a benchmark involving 15 adaptive attempts. As per the blog post, Google would expand access to the model and offer 10,000 scientists six months of free access to the highest usage tier in Google Antigravity, subject to eligibility, capacity and selection. It also said Gemini 4 Argon was the beginning of a wider Gemini 4 model family. (ANI)
|
