Monday afternoon, stuck on “Which monthly plan do I subscribe to?”

At 3 p.m. Monday, I was staring at two invoices. My Claude Code monthly allowance had run out early, and a client was pressing me for script changes. I switched to Gemini, then worried the new environment would throw me off and force me to start over. I’m not a developer serving developers; I’m a freelancer who takes solo projects, writes some code, and also has to deliver documentation. The annoying part isn’t the price. It’s adapting to a different tool every time.

What I compared—and who’s already using each

Google released the fourth-generation Gemini 4 in under two months. Under default settings, the Neon version’s score was essentially level with Claude Code. Raising reasoning strength added just $2.70 per task and earned one more point. I tested both with real client work: an order-cleaning script for a guesthouse, using the same brief. The scores were close, but the prices differed by almost half. That turned the question from “Which is stronger?” into “Which fits my current work?”

I know a peer in Hangzhou, A Jie, who maintains a small WordPress site. He runs his routine changes on Gemini because there are many small jobs and the unit price is lower. Zhou Jing, another freelancer in Chengdu who builds data dashboards, sticks with Claude Code because complex logic passes in one attempt. Less rework means less stress, and for her, that matters more than saving money. If you’re genuinely torn, start by counting how many types of coding work you handle in a week.

What it costs to reproduce this today

Money: Both use subscriptions, with basic plans ranging from roughly $15 to $30 or more per month; higher-volume users can pay by usage. Time: Install, connect your account, and run one of your own tasks—less than an hour. Technical barrier: You don’t need to write code, but you do need to understand a task brief well enough to specify where files belong and other details. First step: Open Gemini’s developer interface and paste in the most urgent small task you have, unchanged. Don’t begin by reading a stack of tutorials. To compare, run the same task on the other tool; I’ve left a comparison template in the table below that you can copy.

Honestly, I also got this wrong the first time. I thought changing the registration bot would be enough, then discovered I also needed to prepare the test environment. I wasted two days.

My pick by team size and type of work

If I’m just starting, with scattered work and no stable clients yet: I’d open only one month of one service, not subscribe to both. I’d get one real job running end to end, see where it breaks, and decide from there.

If I have one or two stable clients and delivery has picked up: I’d split tasks by type—small, repetitive changes on the cheaper side, complex logic that needs to pass on the first attempt on the tool that delivers that. The price difference between the two is often cheaper than the time lost to one round of rework.

If I’m scaling and someone on my team doesn’t code: I’d track completion time and cost for each task type in a spreadsheet, then decide what to expand after a month of data. At that stage, usage-based pricing becomes more worth investigating than a flat subscription.

Not everyone needs these tools. If your current work never touches code, there’s no urgency to try them now. My own answer is simple: test both with one real task, then decide who gets the long-term budget.