Unverified50% confidenceFactExact time
The model places n-gram lookup tables at layers 1 and 14, with 24 hashes each indexing approximately 16 million rows, totaling 196 billion parameters, exceeding one-third of the model's weights.
1
Sources
50%
Confidence
Long-term
Relevance
9/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAnalysis speculates that the model is an MoE model with 2 to 4 trillion parameters, activating only approximately 150 billion parameters per token.77% similarUnverifiedThe GPU resources required for a single inference of a hundred-billion-parameter flagship model may be 10 to 20 times higher than those of a ten-billion-parameter model.74% similarUnverifiedTongyi Qianwen 3.8 has approximately 2 trillion parameters, and Kimi has approximately 2 trillion parameters.73% similarUnverifiedThe final parameter count of the model is 13,693,824, falling within the target range of 10–30 million.73% 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.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/923874API
curl https://kongchang.com/api/v1/knowledge/claims/923874MCP
get_claim(id=923874)