I Dropped Claude Pro and Built a Better AI Stack for $60/Month

I was paying $100 a month for Claude Pro. Then I actually looked at what I was using it for.

Coding tasks. Image generation. Prompt iteration. General back-and-forth.

That's not one tool. That's four different jobs. And no single subscription was doing all of them well.

So I stopped paying for the bundle and started paying for the parts.

The Stack

Here's what I actually pay for now:

Max — $20/month One specific use case that this particular model is genuinely best at. I won't pretend otherwise.

ZDI (GLM) — $20/month For coding tasks. Different model, different strengths. The trade-offs are worth it because I'm not paying for everything else in one place.

MiniMax — $20/month Image generation. This is the part Claude Pro couldn't do at all without an additional subscription anyway. Now it's a first-class part of the setup.

GPT (free tier) Prompt creation. I can get away with the free tier for this because it doesn't need heavy reasoning — it just needs to get out of the way and let me iterate.

Total: $60/month. Less than Claude Pro alone. And I have more limits, not fewer.

Why This Actually Works

The thing about "one model to rule them all" is that it only makes sense if one model is actually best at everything. They're not. Different models are built differently, trained differently, and excel at different things.

When you split tasks across specialised tools, you're not adding complexity — you're matching the tool to the job. That's not fragmentation. That's just using the right instrument.

What I'd Tell Myself Three Months Ago

You don't need the best model for everything. You need the right model for each thing. The $100/month subscription feels safe because it's simple. But simplicity is expensive.

The stack costs less, does more, and I can point to exactly which tool is doing which job.

That's a better setup than one bill and a lot of compromises.