No, that's a compatibility thing after they changed the behavior of the aliases.
Or maybe it was calling `reasoner` instead. Whatever it was, the billing definitely showed 100% DeepSeek V4 Pro usage for the benchmark. My only usage was the benchmark, and all usage was Pro. (I only noticed that there was a problem in what the benchmark was calling because in a later run, I started seeing Flash usage, which wasn't what I wanted to test.)
I'm absolutely confident the benchmark results were using DeepSeek V4 Pro. It would be useful to also have Flash data, but the report I linked is all Pro.
Or maybe it was calling `reasoner` instead. Whatever it was, the billing definitely showed 100% DeepSeek V4 Pro usage for the benchmark. My only usage was the benchmark, and all usage was Pro. (I only noticed that there was a problem in what the benchmark was calling because in a later run, I started seeing Flash usage, which wasn't what I wanted to test.)
I'm absolutely confident the benchmark results were using DeepSeek V4 Pro. It would be useful to also have Flash data, but the report I linked is all Pro.