The smaller models aren’t actually R1, but Qwen/Llama tuned on R1.

Maybe they got poached elsewhere? There was Mistral, but it got bought out by Google.

Their own older model, no less.

It would be weirder/more of note if their new model was worse.

Since those smaller models are technically fine-tunes of Meta/Facebook’s LLAMA, using Deepseek’s outputs, I wonder if they would be covered by the bill at all.