[NotMichaelBay]

I was under the impression that this model does support caching. The pricing page says the cost of input tokens (cache hit) is $0.028.