.png)
News in AI 14-04-2025
- learnwith ai
- Apr 14
- 3 min read
OpenAI is rolling out the full GPT-4.1 model family standard, mini, and nano to all developers, and at significantly reduced costs. These models are faster, smarter, and more cost-effective, with GPT-4.1 nano emerging as the fastest and cheapest option yet.
Here’s how the pricing now stacks up:
Model | Input | Cached Input | Output | Blended Pricing* |
gpt‑4.1 | $2.00 | $0.50 | $8.00 | $1.84 |
gpt‑4.1 mini | $0.40 | $0.10 | $1.60 | $0.42 |
gpt‑4.1 nano | $0.10 | $0.025 | $0.40 | $0.12 |
Developers can also benefit from a boosted prompt caching discount of 75% for repeated queries, and long-context inputs are now supported at no added cost.
These updates aren’t just about making AI more affordable they’re about democratizing advanced language models for everyone from startups to enterprise.
Meta’s AI Training in the EU: Transparent, Local, Human-Centered
Meta is now using publicly shared posts and comments from adults in the European Union to train its AI systems. This bold step is designed to ensure that Meta AI understands the full cultural, linguistic, and regional diversity of Europe.
Key Highlights:
Only public content from adults is used minors’ data and private messages remain untouched.
Clear, accessible opt-out forms are being rolled out to all users in the EU.
The move complies fully with guidance from the EDPB and Ireland’s DPC, after pausing training last year for regulatory clarity.
Why this matters? Meta is aiming for AI that isn’t just built for Europe it’s built with Europe.
From humor and slang to hyper-local references, the company wants its generative models to reflect the rich complexity of European life. Meta joins the ranks of OpenAI and Google in using European public data, but it emphasizes that its rollout is more user-informed and transparent than industry norms.
NVIDIA Brings AI Supercomputing Home: Made in the U.S.
While OpenAI and Meta are reshaping how we use and train AI, NVIDIA is re-engineering where AI lives bringing AI supercomputer manufacturing to American soil for the first time.
The company has commissioned over a million square feet of manufacturing space across Arizona and Texas to build and test its advanced Blackwell chips and AI supercomputers.
Highlights of NVIDIA’s U.S. Expansion:
TSMC is manufacturing Blackwell chips in Phoenix, Arizona.
Foxconn and Wistron are building AI supercomputer factories in Houston and Dallas.
Packaging and testing are handled by Amkor and SPIL in Arizona.
Mass production is scheduled to ramp up within 12–15 months.
Within four years, NVIDIA estimates that it will deliver up to $500 billion worth of AI infrastructure in the U.S., fortifying national supply chains and creating hundreds of thousands of jobs. These “AI factories” are optimized for generative workloads entire data centers designed to power a new era of cognitive computing.
The factories themselves will also leverage AI and robotics, with NVIDIA using Omniverse to create digital twins of its facilities and Isaac GR00T to automate operations.
As CEO Jensen Huang puts it,
“The engines of the world’s AI infrastructure are being built in the United States for the first time.”
The Convergence of Intelligence, Identity, and Infrastructure
What do all three of these moves have in common?
They reflect a maturing AI ecosystem one where cost, context, and compute location matter more than ever.
OpenAI is setting the new benchmark for affordable, scalable intelligence.
Meta is aligning AI to cultural realities, with a region-first approach.
NVIDIA is reshoring the very heart of AI compute, creating a resilient, localized foundation.
In short, 2025 is not about AI expansion it’s about AI alignment. Alignment with developers, with users, and with national strategies for innovation and security.
Resources:
—The LearnWithAI.com Team
