Unverified50% confidenceBenchmarkExact time
After adopting the causal encoder-decoder, the activated parameters per token during the prefill phase were reduced from 16B to 8B.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn speculative decoding, the draft model's parameter count is typically about 1/10 to 1/5 of the target model, and the candidate count k is typically 4-8.69% similarUnverifiedIn the CED architecture, only 8B parameters are activated on the input side, while 16B parameters are activated on the output side.68% similarUnverifiedMiniMax M3 每个 token 推理时约激活 23B 参数66% similarUnverified推理端优化技术包括量化、知识蒸馏和KV Cache，量化可将参数从32位或16位浮点压缩为8位甚至4位整数65% similarUnverifiedEarly testing on the Opus model showed an exact string recognition rate of 0 out of 15, whereas after the release of Fable 5 the same test yielded 13 out of 15.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/900719API
curl https://kongchang.com/api/v1/knowledge/claims/900719MCP
get_claim(id=900719)