Unverified50% confidenceBenchmarkExact time
On the arithmetic control task, all models' accuracy stayed the same but the cost per correct answer rose by 1.65 to 2.5 times.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/12/2026
First Seen
Valid until: 1/10/2027
Sources
Related Entities
Related Claims
UnverifiedIn 7 of the 12 model × task combinations, high only changed the price while accuracy stayed the same, with the cost per correct answer rising by 1.5 to 3.4 times.80% similarUnverifiedIn a simulation of 2000 requests, the small model (1 unit cost/request) had an accuracy of 74.5% and a cost of 1.34 per correct answer; the large model (20 unit cost/request) had an accuracy of 95.8% and a cost of 20.88 per correct answer.77% similarUnverifiedWhen budget and timeline calculations were delegated to an external reliable process rather than inferred by the model itself, accuracy improved by as much as 42 percentage points on weaker models and only 2 to 5 percentage points on stronger models.74% similarUnverifiedThe logical reasoning fix in gpt-5.4-mini caused the cost per correct answer to rise by 8.5 times.74% similarUnverifiedTwo models with similar intelligence scores may differ in inference cost by several times or even an order of magnitude.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1007701API
curl https://kongchang.com/api/v1/knowledge/claims/1007701MCP
get_claim(id=1007701)