Model Rankings

How the models on the Model Router compare on capability, price, and the trade-off between them. Every figure uses live retail prices, with the 5 percent platform fee already included, so the comparisons reflect what you would actually pay.

Intelligence scores come from the Artificial Analysis intelligence index, an independent composite benchmark scored from 0 to 100. Models they do not track are left out of the intelligence-based charts rather than scored as zero.

How to use these charts

Benchmarks measure general capability on standard tasks. Your workload is not a standard task. Use the charts to build a shortlist of two or three candidates at the price point you can afford, then test them on your own prompts and compare the results.

A few patterns worth knowing:

• Price and intelligence are only loosely related. Several mid-priced models score close to models costing ten times more.
• Value scores favour cheap models. A model with a modest score at a very low price can top the value leaderboard while still being the wrong choice for hard work.
• Provider averages hide wide ranges. A provider's cheapest and most expensive models can differ by two orders of magnitude, so the average is a rough orientation only.

For the full price list and how charges are calculated, see the Model Catalog.