How we score
Updated
Short version: every "Numptr says" score is our editorial judgement, out of 10, in half points. It isn't a lab benchmark and it isn't a formula. We read the vendor's own pages, check the prices and free tier on the date shown, look at how the tool is actually being used, and make a call. Then we write down why, in one line, next to the score.
What we weigh
Five things. Every tool gets judged on all of them. The weights show roughly how much each one counts in our thinking. They're a guide we hold ourselves to, not a spreadsheet that spits out the number.
| Criterion | Rough weight | What we're asking |
|---|---|---|
| Does the job well | 40% | For the job this category is about, is it one of the best tools going? Does it hold up on real work, not just the demo? |
| Price and value | 20% | What does a builder actually pay, and is the bill predictable? Surprise overages and confusing credit systems cost points. |
| Free tier | 10% | Can you try it properly before paying? A real free tier beats a seven-day trial; no free option at all costs a little. |
| Momentum and adoption | 15% | Are people genuinely using it, and is it shipping? Fresh launches with no track record score cautiously until they've earned it. |
| Openness and lock-in | 15% | Can you leave? Open source, self-hosting, exportable code and bring-your-own-model all help. Walled gardens don't. |
How scores become ranks
- Within each category, rank follows score: higher score, higher rank.
- When two tools tie, we make an editorial call and the order stays put until something changes. A free tier usually breaks the tie.
- If we can't verify a tool's pricing or it's too new to judge, it shows as "Not scored yet" and sits at the bottom of its category until we can.
What we don't do
- We don't sell scores or ranks. Not to sponsors, not to affiliates, not for a pint. Sponsored spots are labelled and kept out of the rankings.
- Affiliate links don't move anything. Where we have one, it's disclosed, and the tool is scored exactly as it would be without it.
- We don't claim tests we haven't run. We don't publish benchmark numbers. Facts on each tool page (prices, free tiers, plans) come from the vendor's own pages with the date we checked them; anything we couldn't confirm is marked as such.
Keeping it fresh
AI tools change weekly. When pricing, plans or capability shift, we re-check the source and rescore if it's warranted, and the "checked" date on the tool page moves with it. Think a score is wrong? Tell us at [email protected] with a source and we'll have a proper look.