Unverified50% confidenceBenchmarkExact time
In the author's main thread, 2,546.9 million tokens were read from the cache while only 24.6 million were written, a ratio of approximately 103:1.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified35 KB of text is approximately equivalent to 8,000 to 10,000 tokens based on an English-language estimate.68% similarUnverifiedThe 128K character output limit is approximately equivalent to 30,000–40,000 tokens.66% similarUnverifiedThe PagedAttention paper uses OPT-13B in FP16 as an example, where each token requires approximately 800 KB, and 2048 tokens require approximately 1.6 GB of KV Cache.62% similarUnverified128K长度的序列所需的注意力计算量是4K长度序列的1024倍61% similarUnverified6.1 Sol reads an average of approximately 110,000 tokens per request, whereas Sonnet 5 requires approximately 360,000 tokens.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/996128API
curl https://kongchang.com/api/v1/knowledge/claims/996128MCP
get_claim(id=996128)