Illustration 01
Google launches Gemini 4 Carbon, a frontier model focused on cybersecurity
Google unveiled Gemini 4 Carbon, a frontier model optimized for cybersecurity and coding tasks, as it competes with OpenAI and Anthropic. Google says that on its internal benchmarks the model produces secure code at more than twice the success rate of its predecessor, Gemini 4 Argon. At first, access is limited to a group of cybersecurity defenders through Google's Fairwind Program while it keeps testing safety safeguards.
Introductory pricing is $2 per million input tokens and $10 per million output tokens.
Illustration 02
AMD to invest $1 billion in Meta and co-design future AI chips
AMD announced a $1 billion investment in Meta's equity offering as part of a wider AI infrastructure partnership. According to AMD, Meta has deployed millions of EPYC processors, runs Instinct GPUs in production for training, ranking and recommendation, and is testing AMD Helios rack-scale systems. AMD provided day-one software support for Meta's Muse Spark Open models on Instinct GPUs.
The two companies are co-designing future platforms, including MI500 series GPUs and a custom Instinct accelerator for Meta.
Illustration 03
Oracle puts Nvidia, AMD, Intel and Google AI accelerators on one cloud platform
Oracle Cloud Infrastructure announced it will offer AI accelerators from Nvidia, AMD, Intel and Google, linked by its own Acceleron fabric. Nvidia Vera Rubin NVL72 systems are generally available now, and AMD MI455X, Intel Diamond Rapids and Google TPUs are planned. Oracle says this will make OCI the first hyperscaler other than Google to deploy TPUs, with its first Cerebras CS3 deployments planned for 2027.
The company also said it added 850MW of data center capacity in the first quarter of fiscal 2027.
Illustration 04
JAMA study links Medicare AI prior authorization pilot to fewer spinal injections
A University of Minnesota study published in JAMA linked Medicare's AI-assisted prior authorization pilot to a 15 percent drop in spinal steroid injections across six states. The study is an early look at how AI used in coverage decisions affects the care that Medicare patients receive.
Illustration 05
MiniMax open-sources MCode Stack for AI coding agents
Chinese AI developer MiniMax open-sourced its MCode Stack, a set of tools that lets engineering teams automate coding tasks with AI agents. Because the tools are open source, any developer can use and adapt them, adding to the open options for agent-based software development.
