Unverified50% confidenceFactExact time
MemFold fixedly uses K=256 continuous vectors, so the model's input footprint does not increase regardless of whether the original history is 32K or 128K tokens.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedA 200K token context request requires approximately 25× the KV-Cache memory compared to an 8K token request.70% similarVerifiedTaking the LLaMA-2 70B model as an example, a sequence of 2048 tokens requires approximately 4 GB of VRAM for the KV cache alone.69% similarUnverifiedModel weights cannot fill up VRAM entirely, as space must be reserved for the KV cache and inference processes; the KV cache required for a 128K context can reach as high as several GB or even exceed the model weights themselves.68% similarUnverifiedWhen the context expands from 4K tokens to 128K tokens (a 32× increase), the theoretical computation grows by approximately 1024×.66% similarUnverifiedFor the LLaMA-2 70B model, when processing a context of 4096 tokens, the KV Cache requires approximately 20 GB of system memory at FP16 precision.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1003152API
curl https://kongchang.com/api/v1/knowledge/claims/1003152MCP
get_claim(id=1003152)