Verified70% confidenceFactExact time
The computational load of the attention mechanism is proportional to the square of the sequence length, meaning doubling the context increases computation fourfold.
4
Sources
70%
Confidence
Long-term
Relevance
7/13/2026
First Seen
Sources
智谱GLM-5.2发布：国产开源模型首次跨入主力门槛
bilibili冶承w6/22/2026
Related Claims
UnverifiedThe computational complexity of the standard self-attention mechanism is proportional to the square of the context length (O(n²)), meaning extending the context from 100K to 1M theoretically increases computation by 100 times.78% similarUnverifiedTransformer注意力机制的计算量随序列长度平方增长，早期语言模型上下文窗口上限约4000个token75% similarUnverified标准Transformer自注意力计算量随序列长度平方增长，50万token的注意力计算量约为4096token的15000倍72% similarUnverifiedOutput token pricing is typically 2–4 times that of input tokens, because the generation process requires step-by-step decoding and computation of attention weights.71% similarUnverifiedCognitive load theory states that human working memory can simultaneously process approximately 4±1 information units.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/501410API
curl https://kongchang.com/api/v1/knowledge/claims/501410MCP
get_claim(id=501410)