比较模型身份与能力。
meta/llama-4-maverick-17b-instruct
Meta / Llama
Open multimodal Llama for strong reasoning with efficient everyday serving