Unverified50% confidenceFactExact time
35B-A3B混合专家模型总参数35B，每次推理实际激活约30亿参数，适合16GB显存的卡运行
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/15/2026
First Seen
Valid until: 10/13/2026
Sources
Related Claims
Verified一个原本需要60GB显存的30B参数模型，经过量化后能在16GB甚至8GB显存的消费级显卡上运行77% similarUnverified35B模型在FP16精度下需要约70GB显存，远超RTX 5090的32GB容量，因此单卡部署需量化74% similarUnverifiedINT4 量化模型有望在 8GB 乃至更低显存的显卡上完成原本需要 16GB 或 24GB 显存的模型推理73% similarUnverifiedLLM推理/训练选卡看显存优先于算力：7B模型微调至少需24GB显存（RTX 4090/A5000），13B需A100 40GB，70B全参数训练需A100 80GB×4或H100集群73% similarVerified阿里巴巴推出Qwen 3.6 35B A3B，拥有350亿参数的混合专家模型，同一时间只有30亿参数处于激活状态，权重约72GB已在Hugging Face开源70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/518291API
curl https://kongchang.com/api/v1/knowledge/claims/518291MCP
get_claim(id=518291)