What our worth-it is based on?

How Highworth turns real community feedback into a simple verdict: worth it, or not.

1. The worth-it score

Every model gets one number: the share of reviewers who would pay for it again. We calculate it as worth-it ÷ (worth-it + not-worth-it), shown as a percentage. A score of 70% means 7 out of 10 people who tested it say it's worth the money.

2. Where reviews come from

We collect real user sentiment from public communities — Reddit, Hacker News, V2EX, Zhihu, X, YouTube, GitHub, Hugging Face and more — and from reviews submitted on Highworth itself. Each quote is attributed to its source (with a link where available) and tagged positive, negative, or mixed. Non-English reviews are translated faithfully before being stored.

We never fabricate reviews. If we can't find real feedback for a model, that model simply shows fewer reviews rather than made-up ones.

3. Prices & versions

Prices are taken from official pricing pages and checked weekly. Models are renamed to their current version, and every page shows the last time the model was updated. When a price or version is uncertain, we mark it rather than guess.

4. This week vs. all time

The homepage “Hot This Week” reflects this week's reviews. Model pages and the model directory reflect the cumulative, all-time score — the long-term consensus, not a single week's noise.

5. The value-for-money index

The worth-it index ranks models by a combination of their worth-it percentage and a price bonus — free and very cheap models get a small boost, so a high score reflects both community approval and reasonable cost.

6. Independence

Highworth is independent. We are not paid to change a score, and sponsored content — if it ever appears — will be clearly labelled and kept separate from the ratings. If revenue ever influenced a ranking, the whole thing would be worthless, and we know it.