Unverified50% confidenceFactExact time
For the same passage, the price when the model reads it the second time is only one-fiftieth of the first time (that is, the cache-hit price is far lower than the cache-miss price).
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/6/2026
First Seen
Valid until: 2/4/2027
Sources
4 Must-Have Plugins for Saving Tokens in DSH — Stop Installing Blindly
bilibili天才依山尽10/29/2026
Related Entities
Related Claims
UnverifiedCognition states that SWE-2 scores only one point below Anthropic's Claude model on its own coding benchmarks, yet costs 64% less65% similarUnverifiedIn this cost breakdown, model replies account for 25.5% of total cost, ranking second.64% similarUnverifiedIf a model nominally priced at half the cost has low reasoning efficiency, requires a lot of code rework, and needs more rounds to converge, the final bill may match or even exceed that of a flagship model.63% similarUnverifiedFor a premium reasoning model costing $0.015 per query, a $0.00002 embedding call corresponds to an economic break-even hit rate below 1%; when routing to lightweight models (such as GPT-4o-mini or Claude 3.5 Haiku) with an inference call cost of $0.0002, an application needs a 15% to 20% cache hit rate to avoid losing money.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1142144API
curl https://kongchang.com/api/v1/knowledge/claims/1142144MCP
get_claim(id=1142144)