Cheaper Than DeepSeek? China’s Z.ai Just Flipped the Script on AI Costs

AI just got way more interesting… and a whole lot cheaper.

Z.ai — a Chinese startup you’ll soon hear everywhere — just launched GLM-4.5, a compact AI model that takes direct aim at DeepSeek… and lands a hit.

How?
It’s half the size, runs on just 8 Nvidia H20 chips, and charges less than half the output cost of its rivals.

Let that sink in.
No 100-GPU setups. No $100M training runs. Just lean compute + agentic intelligence = smarter tasks, faster, and cheaper.

And yes — it’s open source.
Free to download. Free to experiment.

While Silicon Valley argues about AGI timelines, China is quietly building models that run leaner, cost less, and still perform hard tasks like code generation.

This isn’t just an upgrade — it’s a shift in who controls AI infrastructure and who can afford to innovate.

Z.ai isn’t alone. Tencent, Alibaba, and others are dropping new models weekly.
China’s playbook? Scale fast, open source it, and make it accessible.

Your move, OpenAI.

If you’re building or curious about open-source AI tools that don’t break the bank — keep your eye on GLM-4.5.

We’re watching the next phase of AI infrastructure play out — and it’s not where you think.