Imagine you’ve been the best student in your class for years. Then one day a new kid transfers in, and by the end of the first week, they’re almost catching up to you — using textbooks that anyone can borrow for free. That’s a little like what just happened in the world of AI.
The “new kid” is called GLM-5.2. It’s a new AI model built by a Chinese company called Z.ai, and it just did something surprising: in coding contests where real people compare AI answers side by side without knowing which AI made which answer, GLM-5.2 came out on top of every other openly available AI model in the world. It’s not quite beating the very best models from companies like Anthropic yet — it’s still a few steps behind on the hardest tasks — but the gap is small, and it’s closing fast.
Think of it like a really dedicated student
Here’s a simple way to picture how an AI model works. An AI model is like a student who learns by studying millions of practice examples instead of reading just a few textbooks. That studying is called “training,” and the finished student — ready to answer your questions — is the “model.” GLM-5.2 studied hard, using roughly 750 billion tiny brain-like connections (engineers call these “parameters”). It can also hold about a million words of conversation in its memory at once, kind of like a student who never forgets what you told them ten pages ago.
Now here’s the part that really matters: GLM-5.2 costs about one-sixth as much to use as the top U.S. models, and it was released for free under an open license. That’s like a straight-A student sharing their entire notebook of study notes with the whole school, instead of keeping it locked away. Anyone — a student, a teacher, a small business owner, a kid tinkering in their bedroom — can download it, use it, and even build their own tools on top of it, without paying anyone for permission.
Why should you care?
Because it means powerful technology is becoming less like an expensive private tutor and more like a public library — available to almost anyone, almost anywhere. That could help a teacher build a free study tool, a small business automate paperwork, or a curious kid build their first app. It also means the AI “race” between countries is getting closer, which is worth paying attention to.
A few questions worth sitting with:
- If two students both earn an A, does it matter that one studied with expensive private tutors and the other used free library books?
- Would you trust a new, cheaper AI tool as much as a famous one? What would you want to check first?
- If powerful AI becomes free for almost anyone to use, what responsibility comes along with that freedom?
- As AI tools get smarter and cheaper every year, what human skill should you practice now so you stay valuable no matter what?
Josiah here: I’ve actually tried this myself. I run an open-source model called Qwen locally on my own PC. The tradeoff is real — since it depends entirely on my computer’s own power, replies take longer than a cloud AI would. But the upside is real too: everything stays fully private, nothing leaves my machine. Someday I’d like to go further and fine-tune and modify a model like this myself, and integrate it into my own projects.
Here’s the practical lesson: don’t judge a tool only by its name or its price tag. The most useful habit in an AI-filled world isn’t picking “the best” model once — it’s staying curious enough to keep comparing, testing, and asking good questions as the tools keep changing.
So next time you use an AI chatbot for homework, work, or just for fun, try asking yourself: am I just trusting this because it’s popular, or because I actually checked that it’s good? That one habit — checking instead of just trusting — might be the most important skill of all.
JOSIAH Miri 