I mean the pricing is high but this is some kind of model where this pricing actually makes sense. Before that the pricing of, let's say, Opus 4.8 or 5 or 6 was very much similar to the Chinese models benchmark but very costly inference. This model is so better in benchmarks that the inference cost actually makes sense.