About Research Work With Me AI News Contact
CLAUDE CODE

What 'Free Claude Code' Actually Means

A post going around says you can run Claude Code for free. It's technically true — and it leaves out the one thing you'd actually want to know.

The Claude Code window with the model name, Opus, circled at the bottom.
Claude Code has two parts: the interface that does the work, and the model — Opus — that decides what to write. Screenshot

There's a post going around right now — maybe you've seen it. It says you can run Claude Code for free: install a free tool, paste in a couple of settings, and you're done. No subscription, no bill. All of it is technically true. And all of it leaves out the one thing you'd actually want to know.

Claude Code is two things

Most people think of Claude Code as one product. It isn't — it's two, and they come apart. There's the shell: the part on your screen. It opens your files, writes the changes, runs the commands, and shows you what happened. That's the part that does the work. And there's the brain: the model deciding what to write. Look at the bottom of the window, where it says Opus. That's the brain.

The free setup keeps the body. And swaps the mind.

The same Claude Code window with a note to swap Opus for Qwen, GLM, or Kimi
Same shell. Swap that one word and you're running something else.

So what are you actually running?

Qwen. Or GLM. Or Kimi. Good models — genuinely. But not Claude, and not close to Opus on the hard problems. You kept the car and replaced the engine. That's not free Claude Code; it's Claude Code's dashboard with somebody else's model underneath.

And to be fair to Ollama — the company behind this — they never claimed otherwise. Their documentation is clear about exactly what it does.

Ollama's official page stating it connects Claude Code to local and cloud models
Ollama's own page says it plainly.

Three things the videos skip

  1. It's not a hack. One command runs the whole setup. It's documented and supported. Nothing shady here — the problem is the headline, not the tool.
  2. "Local" often isn't local. Ollama runs models on your machine, but it also runs them in the cloud — and a lot of those demos use the cloud version. That means your code is still leaving your computer, and the free tier runs out. Two very different things wearing the same word.
  3. Free still costs you. Running a model on your own machine takes a real machine — around 16GB of memory before it gets painful. And the file-editing you were excited about only works with certain models; pick the wrong one and the shell you kept starts failing at the exact thing you kept it for.

Weaker brain, weaker everything.

You kept the car. You replaced the engine.

Dr. Erin Jacques

Would I use it?

If I hit my usage limit in the middle of something? Yes. It's a real option, and I'd rather you know it exists than not. But I'd know what I traded. That's the actual skill here — not memorizing tools, but understanding what a tool is made of, so nobody can hand you a swapped part and call it the original.

That's the whole game inside my community, Leveraging AI — understand how these tools really work, then use them to build AI web apps people pay to subscribe to. No code required.

Dr. Erin Jacques

Dr. Erin Jacques is a professor of informatics at the City University of New York and the founder of Leveraging AI and ChatifyIT, where she helps people build and monetize AI-powered web apps people pay to subscribe to.