Welcome to Tech Beat, your daily read on the technology stories that matter.
The build versus buy debate in AI just got a fresh data point. A hardware reviewer pitted a dual AMD Radeon AI PRO R nine seven hundred workstation against OpenAI's cloud pricing, and the numbers are striking. At roughly eighteen thousand seven hundred seventy five dollars upfront, the rig pays for itself when a team burns through twenty million tokens a month, saving nearly eleven thousand seven hundred dollars annually. Multi-token prediction pushed throughput to over three hundred twenty tokens per second, which is where, as one analyst put it, the machine starts winning on cost.
Speaking of OpenAI, the company has begun rolling out GPT-six Astra to users, and early guides are already circulating on how to access it and what pitfalls to watch for. The release follows a rapid succession of model updates, and it signals that the pace of frontier model deployment is not slowing down anytime soon.
Shifting to a story that sits at the intersection of intelligence work and personal finance, a former CIA official accused of hoarding forty million dollars in gold bars has reached a tentative plea deal. The case raises uncomfortable questions about financial oversight within agencies that handle some of the country's most sensitive work.
That is the shape of today's tech landscape. Keep surfing. Tech Beat out.
