Unverified50% confidenceFactExact time
estimateRequestInputTokens and createTokenEstimator no longer compute base64 image payloads at the text rate; the new version adds a fixed constant ESTIMATED_TOKENS_PER_IMAGE (1600) for each image.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/13/2026
First Seen
Valid until: 1/11/2027
Sources
Related Entities
Related Claims
UnverifiedIn the old version, the client would crash directly due to image decoding errors when processing images; this update fixes the issue so the client can read image content normally.63% similarUnverifiedImages 2.5 can retain the record of previous changes across multiple rounds of editing, and the stability of portrait features is superior to version 2.0.56% similarUnverifiedThe idempotency design writes a marker TX_DONE_{youtubeId}_{lang} before initiating the request, and skips if the marker already exists, to avoid duplicate charges caused by retries after a timeout.53% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1016324API
curl https://kongchang.com/api/v1/knowledge/claims/1016324MCP
get_claim(id=1016324)