Good catch. If it's "too slow" even when ran in a state-of-the-art datacenter environment, this "Mythos" model is most closely comparable to the "Deep Research" modes for GPT and Gemini, which Claude formerly lacked any direct equivalent for.