Unverified50% confidenceFactExact time
根据DeepMind在2022年提出的Chinchilla缩放定律，一个700亿参数的模型需要约1.4万亿token才能达到计算最优
1
Sources
50%
Confidence
Long-term
Relevance
8/20/2026
First Seen
Sources
Related Claims
UnverifiedThe Chinchilla optimal scaling strategy proposed by DeepMind states that for a given compute budget, there is an optimal ratio of parameter count to data volume, approximately 1:20.73% similarUnverifiedAnalysis speculates that the model is an MoE model with 2 to 4 trillion parameters, activating only approximately 150 billion parameters per token.70% similarUnverifiedA model with 100 billion parameters requires approximately 80 GB of VRAM to process a single inference request, while a model with tens of billions of parameters requires only about 20 GB.69% similarUnverifiedMoE架构中，一个标称800亿参数的模型每次推理时可能只激活约200亿参数，VRAM占用大幅降低69% similarUnverifiedAlibaba's new-generation Qianwen model has 2.4 trillion parameters, making it the first multimodal large model in the series to exceed one trillion parameters; its code generation accuracy is reportedly 94.7%, an improvement of approximately 11 percentage points over the previous generation.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/778687API
curl https://kongchang.com/api/v1/knowledge/claims/778687MCP
get_claim(id=778687)