Model comparison

Benchmark score and estimated no-cache USD per 1M total tokens

Report generated:

Pricing data:

Billing
Pricing basis
X-axis scale
Plans used for estimates

Toggle plans individually; the lowest enabled matching plan is used for each model.

Score × unit price

Hollow circles show the estimated API value used when no enabled subscription plan matches

Results

Displayed values are estimated USD per 1M total tokens. Filtered by the current settings, sorted by score (highest first)

Model Agent Effort Score Estimated API USD / 1M total tokens Selected estimate USD / 1M total tokens Rate basis

How to read this report

API unit rates are no-cache estimates in USD per 1M total tokens: (5 × input price + output price) ÷ 6, a normalized 5:1 input-to-output mix. Cached input and cache-write prices are excluded. Components are selected in this order: valid metered input/output, hand-maintained official input/output, then valid components from other source rows. These estimates are not the original observed API costs.

Subscription rates are estimates: each source subscription rate is scaled by the ratio of the new no-cache API rate to its legacy API reference rate. This assumes the monthly allowance is fully used and the subscription discount rate remains unchanged; it is not observed subscription spend. For rollouts, reported tokens are multiplied by the estimated rate; otherwise, original API cost is used only to infer equivalent token volume. Missing inputs show —, without API fallback.

Compare scores within each benchmark only. Scores from different benchmarks are not combined or directly comparable.

Data sources