Unverified50% confidenceFactExact time
v4.1 Flash introduces a mechanism called Ngram, where 2 out of 40 layers contain conditional memory tables, with 196 billion parameters used for a lookup table that matches sequences of up to 4 tokens against a vocabulary of 16 million.
1
Sources
50%
Confidence
Long-term
Relevance
9/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedQwen3.8 Flash Next additionally introduces an Ngram module with 51 billion parameters, resembling a massive local phrase memory table.80% similarUnverifiedV4.1 Flash activates 8 billion parameters when processing input and activates 16 billion parameters when generating output.75% similarUnverifiedThis evaluation used six composite questions covering front-end, reasoning, and causality, with V4.1 Flash, V4 Flash, V4 Pro with deep thinking enabled, and JLM 5.3 Flash tested back-to-back on the same questions.71% similarUnverifiedv4.1 Flash uses a Mixture-of-Experts (MoE) architecture, with each layer containing 384 routed expert networks plus 1 shared expert, and the router selects 6 of them for each token.70% similarUnverifiedIn the vector design tool and state machine logic question, all four models passed, with V4.1 Flash performing most stably and achieving the best industrial UI aesthetic.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/923763API
curl https://kongchang.com/api/v1/knowledge/claims/923763MCP
get_claim(id=923763)