⚡Megatron LLM Hub
←返回搜索
Twitter

观点:单张 RTX 3090 上最好的本地模型是 Qwen 3.8 27B dense

自托管 agent 的模型与硬件选型参考,决定本地推理门槛
llm本地部署
RT @sudoingX: in case you forgot, the best local ai on a single rtx 3090 is qwen 3.8 27b dense and it's not even close