Provider-level cache hit rates and Cost per Task data go live on Artificial Analysis

Artificial Analysis says it now publishes measured cache hit rates for the model endpoints it tracks and provider-level Cost per Task for open-weights models. For GLM-5.3 (max), its calculated Cost per Task ranges from $0.92 on Makora to $8.68 on Nebius. It attributes the roughly ninefold gap to input pricing: both providers cache nearly all eligible input tokens, but Nebius does not discount cached input on its shared serverless endpoints.