How Inferse ranks

Model lists are sorted by the figure in their title — price per million input tokens, context length or release date — exactly as each provider publishes it. Platforms are ranked on what each vendor publishes: an API, a free tier, open code and documentation depth. Paid placements never change a rank.

What goes into a score

Every ranked list on Inferse uses the same published signals, read from each maker's own pages. A score runs from 4.0 to 9.8:

• 38% — how established the product is in its category (makers that buyers already compare)
• 22% — how completely the maker documents its product, plans and terms
• 16% — whether there is a public API
• 12% — whether there is a free plan
• 12% — whether the code is open source

Lists that are directories rather than rankings — sorted by year, by code or by a single published figure — say so in their title and are never called "best".

What never changes a rank

• Payment. Sponsored rows are labelled and sit beside the ranking, not inside it.
• A claim. Makers can correct facts; they cannot set their score.
• Affiliate links. Whether a link earns anything has no part in the order.

Freshness

Each list shows the date its facts were last checked. Facts are re-read from the source when a maker changes a page or sends a correction.