An $80-a-Month Coding Tool MatchedWhat I Was Paying $200 a Month For

For one weekend I ran an open-weight coding model against the two frontier tools I normally pay for, on a real task. It kept up, it used a fraction of the tokens, and it defaulted to better privacy than either of the incumbents. It also has one real gap, which I am naming rather than glossing over.

The problem: renting a week of tokens in an afternoon

A week ago I wrote about hitting a weekly usage cap on Claude Fable 5.1 and watching GPT's Astra burn a week of tokens in one long night (Sort the Task, Then Pick the Model). Astra and Fable are the two models I reach for when I want to do something I could not do before: research an idea nobody has tried yet, or push through something hard that a weaker model would have talked me out of. The tradeoff has never changed. Both are rationed, and it does not take much to burn a week's worth of tokens by the afternoon.

What I tried, and why I doubted it

So this weekend I tried the thing a long-time Claude Code and Codex user is not supposed to try: an open model called GLM-5.3, built by a company named Z.ai, running through their own coding app, ZCode. I have used Claude Code for years and Codex more recently, and I assumed a smaller open model had to be cutting a corner somewhere I had not found yet.

I gave it a business idea I had been sitting on for months. It handled the problem the way I expect Fable or Astra to handle it, not the way I expect a cheaper model to.

Three things surprised me enough to write this up.

The switching cost is close to nothing

The first surprise was how little it cost to try. ZCode ships with a migration wizard that imports your conversation history straight out of Claude Code, along with your connectors and skills (ZCode install docs). I did not have to rebuild my setup to see whether the model itself held up. For a business evaluating a second AI vendor, that is the number that actually matters: not whether the new tool is good, but whether trying it costs you a rebuild.

The math: $80 a month versus $200

The second surprise was the price. I am on ZCode's Pro plan, $80 a month, less if you pay quarterly or annually. A multi-hour coding session that would have burned something like 30 percent of a $200-a-month Claude Max or ChatGPT Pro week used about 6 percent of my ZCode week for the same kind of work.

That is not a claim that GLM-5.3 beats Fable or Astra across the board. It is one weekend, one task. But it is a real enough gap that any business paying for a top-tier coding subscription should at least be pricing out whether a cheaper second option covers some of that work.

The privacy default nobody expects from the challenger

The third surprise is the one I did not see coming. I checked the setting that controls whether the vendor trains on your conversations by default, the way I check it on every new AI tool. It was off. The exact wording: "Allow us to use your conversations to improve the Agent experience. We protect your data privacy and security," unchecked, no action required from me.

Anthropic's consumer plans, Claude Code included, defaulted that same setting to on in August 2025 unless you actively opted out (Anthropic: Updates to Consumer Terms and Privacy Policy). OpenAI's ChatGPT and Codex still default "Improve the model for everyone" to on today, and that setting covers Codex tasks on a personal plan too (OpenAI Data Controls FAQ). The smaller challenger shipped privacy-first. The two incumbents did not.

If your business runs any AI coding tool on a personal or team plan, this is a five-minute check worth doing today, regardless of which vendor you use: find the training-data toggle, and confirm which way it defaults.

The one real gap

None of this makes ZCode a clean sweep. There is no zero-data-retention option on this plan. Zero data retention means the vendor's systems do not keep a copy of your prompts and outputs beyond what is needed to serve the request, and it typically only comes with an enterprise agreement rather than an individual or Pro-tier plan. If your work legally requires it, whether for healthcare, legal, or financial client data, ZCode is not there yet. Weigh that against the training-default advantage above: a privacy-friendly default is not the same guarantee as a contractual zero-retention commitment, and a business handling regulated data needs the latter, not just the former.

ZCode also offers computer control and browser control, which I have not tested yet, so I am not making a claim about either one.

The takeaway

I went into this weekend expecting a cheaper knockoff and a reason to go back to what I know. I did not find either. What I found instead is a real second option, cheap enough to run alongside a frontier subscription rather than instead of it, with one clear disqualifier for regulated work.

Have you ever swapped out the tool you normally reach for and been surprised by what came back? I would like to hear it.