Unverified50% confidenceBenchmarkExact time
Matched profiles generated 1.5 to 2.3 times the number of output tokens compared to the baseline, and the cost per successful call was 2.2 to 4.5 times higher than the baseline.
1
Sources
50%
Confidence
Long-term
Relevance
10/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn real-world data, the manual cost of pixel-level annotation can reach several dollars or even tens of dollars per image64% similarUnverifiedThe reason multimodal input costs are high is that models need to convert images or PDFs into hundreds to tens of thousands of tokens through a visual encoder (such as ViT architecture), and inference costs are approximately linearly correlated with the number of input tokens.63% similarUnverifiedGenerating a 30-second 720P video directly with Seedance 2.5 costs approximately 56.7 yuan, and using a reference video raises that to 68.4 yuan, an increase of about 20%.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/966776API
curl https://kongchang.com/api/v1/knowledge/claims/966776MCP
get_claim(id=966776)