Expired50% confidenceBenchmarkExact time
In the blogger's sample, approximately 94% of Astra's input and approximately 79% of Sonnet's input came from cache.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
9/16/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 9/30/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedWhen Sonnet's context exceeds the 50,000 to 100,000 token range, its accuracy in locating repeated words within the context window drops by nearly 50%.66% similarUnverifiedThe poster noted that cache hit rates in coding tasks are typically as high as 90% or more.62% similarUnverifiedFor every 10% increase in the draft token acceptance rate of speculative decoding, overall throughput can improve by 15%–25%.59% similarVerifiedThe price of output tokens is typically 4–6 times that of input tokens, stemming from the computational differences between the prefill and decode phases, with GPU utilization during the decode phase potentially being only 10–20% of that during prefill59% similarUnverifiedIn a test-paper grading case, Cowork generated an analysis and HTML report containing an average score of approximately 72%, a minimum of 30 points, and a maximum of 100 points through a single instruction.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/924288API
curl https://kongchang.com/api/v1/knowledge/claims/924288MCP
get_claim(id=924288)