It is also the most expensive open source frontier model, per task; cf. Cost per Intelligence Index Task. If it is as good as the benchmarks indicate it bodes well for Qwen and China. For my part, I'll pass; it is not on the Pareto frontier.