My AI bill last month made me stop cold
I opened my Claude billing page at the end of last month and just stared. My costs had jumped almost 50% from the month before. I hadn't done anything differently — I'd just switched the default model from the older version to the newer one. That single change pushed my bill up more than 40%. My first instinct was: did I use it way more? I checked. Usage was basically identical. The model was the expensive part, not me.
If you're using AI tools to write copy, organize client notes, or draft proposals, this is worth five minutes of your time.
What's actually going on — and someone already ran the numbers
An indie developer named Bill Chambers built a page that tracks exactly how many tokens (think: the words the AI reads and writes — more words, higher cost) different models consume for the same task. His latest data shows Claude Opus 4.7 costs roughly 45% more than 4.6 — and most people never noticed , because there was no alert in the interface. It just quietly switched.
A friend of mine — Xi aoshan, a independent consultant based in Shenzhen — uses AI every morning to process the previous day's client conversations and draft follow-up emails. She told me she only realized last month that she'd been on the most expensive version the whole time. "I just assumed expensive meant better," she said. "Then I switched back to the cheaper one and honestly couldn 't tell the difference."
Here's the thing: for most everyday tasks, the cheaper model is already enough. The premium version earns its price on genuinely complex reasoning — writing code, running data analysis. But if your work is mostly writing copy, organizing information, and producing first drafts, the lighter version handles it fine.
How to check and where to cut costs today
Money involved : If you're on a pay-per-use AI service, the price gap is real. If you're on a flat monthly subscription, this won't hit you directly right now — but it's still good to know.
Time required: Checking which version you're currently on takes under five minutes.
Technical barrier: Zero. You're just looking for a "Model" or "Model Selection" option in the settings page and flipping a switch.
First step: Open whatever AI tool you use most — Claude, ChatGPT, or something similar — and find the "Model" option in the top-right corner or settings. See which version you're on. If there 's a "Lite," "Standard," or similar option, try running one of your usual tasks on the lighter version and see if the output works for you.
This isn't some advanced move. I only looked into it because my invoice forced me to.
Where you are right now, and what I'd do in your position
If you're just getting started with AI tools and still figuring things out: I'd say use whatever version the platform gives you by default and don't overthink it. Once you've settled into a workflow and know what you're actually using AI for, that's the right time to revisit this. Ignoring it for now is completely fine.
If you've got one or two steady clients and you're using AI weekly: It 's worth spending one session comparing the cheaper and pricier versions on your real tasks — writing proposals, replying to emails, whatever you do regularly. When I tested this myself, the lighter version handled about 80% of my daily work just fine. Your situation might be different, so make your own call — don't just copy mine.
If you're scaling up, your team is starting to use AI together, or costs are a number you're actively watching: That 45% gap deserves serious attention. Set aside an afternoon, run your team's most common task types through different model versions, log the results, and decide on a default. The savings from that one session might surprise you.
There's no universal right answer for which version to use. I just figure: if you can get the same result for nearly half the price, it's worth knowing that option exists .