Unverified60% confidenceFactExact time
Google's Switch Transformer, Mistral's Mixtral 8x7B, and the reportedly GPT-4 architecture all adopt sparse MoE designs, activating only a subset of parameters during inference.
2
Sources
60%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
Related Claims
VerifiedModern AI systems such as GPT, BERT, and ChatGPT all have the Transformer as their core architecture.72% similarUnverifiedGPT-6 Astra is speculated to use a recurrent Transformer architecture, meaning it performs multiple forward passes before outputting a single token.71% similarVerifiedThe GPT series uses the Decoder-only variant of the Transformer, predicting the next Token autoregressively.71% similarUnverifiedThe Transformer architecture was proposed by Google in 2017, and ChatGPT is based on the GPT architecture and aligned with human preferences through RLHF.70% similarUnverifiedGPT系列、LLaMA、Gemini等主流模型均基于Transformer的解码器（Decoder-only）变体70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/493409API
curl https://kongchang.com/api/v1/knowledge/claims/493409MCP
get_claim(id=493409)