Cheaper models, big checks and a very public IPO filing. Here's your Tuesday AI rundown.
OpenAI released GPT-6.1 Terra, a faster, cheaper model built for computer work. The company says standard API prices are half what GPT-5.6 Terra costs, and that its mistake rate on a low-effort error test fell sharply. Those are OpenAI's own benchmarks, so treat them as claims until outsiders test them. One flag: OpenAI says Terra meets the "Critical" cybersecurity threshold under its Preparedness Framework. For context, see our earlier look at how OpenAI moved frontier training offline.
Nvidia (NVDA) will invest $3 billion in an equity offering by Meta (META), which makes Nvidia a buyer of new Meta securities rather than a lender or a supplier alone. We couldn't confirm the terms, such as the share price or the exact type of security. The investment sits beside a deeper working relationship on Meta's Muse models, and Meta plans to run Nvidia's newest CPUs and networking gear in its data centers. That is the likely motive for Nvidia: a customer that is also a shareholder is a customer with reasons to keep buying its hardware. The pitch is AI agents, software that does tasks for you, running more cheaply per watt. That's a lot of trust between two companies that also compete for the same talent.
IIT Kharagpur plans to open India's first school devoted to physical AI. That's AI that senses and acts in the real world, so think robots, sensors and digital twins rather than chatbots. The Rs 35 crore project (roughly US$4 million) is funded by the Vinod Gupta family, an anonymous alumnus and the institute itself. Classes are expected next academic session, subject to approvals, and director Suman Chakraborty says he wants AI that isn't a black box. Related reading: Texas Instruments puts $500 to $1,000 of chips in each humanoid robot.
Anthropic has filed for an IPO, and the prospectus is being picked apart. We haven't read the filing ourselves, so we won't repeat figures we haven't verified. Our earlier story has the background: Anthropic eyes November IPO with $16B in sales and $600B in commitments.
Nvidia also named two tools meant to keep AI agents on a leash: OpenShell for deploying them and Sentry for monitoring them. Meta only said it will "consider" using them, so this is an offer, not a deployment. It matters because agents that act on your behalf need guardrails before they need speed. We covered the push in Nvidia wants AI agent safety rules enforced in the chip itself.
Which of these will matter most to your own work or wallet this year?
The WattsUpNext take: AI is getting cheaper to run, and the money behind it is getting bigger and more public. Safety tools are arriving, but so far mostly as offers rather than rules.
That's wattsup, and here's wattsnext: ChargePoint's Express Solo hits 750 kW, but sharing power is the point
Sources
- 1.NVIDIA and Meta to Expand Collaboration on Muse and Muse Family of Models Amid Robust Growth · NVIDIA Newsroom
- 2.Introducing GPT-6.1 Terra · OpenAI
- 3.IIT Kharagpur to Build India’s First Physical AI School: What Makes It Different? · APAC Digital News Network
- 4.IIT Kharagpur Announces India's First School for Physical AI - Check Courses, Curriculum Details · Telegraph India
- 5.In A First, IIT Kharagpur Plans To Unveil Rs 35 Crore Physical AI School · www.ndtvprofit.com
Reported by the WattsUpNext desk from the sources linked below. Spot an error? Tell us at corrections@wattsupnext.com.
The WattsUpNext Brief
Get stories like this in your inbox.
One email each weekday, only the topics you choose. Real news, sourced, no fluff. Unsubscribe in one click.
