Unverified50% confidenceFactExact time
The original methodology specifies: list prices are read from developers' official pricing pages, percentages are rounded down, and 0% means flat or higher than the list price; dated snapshot versions and different effort variants are merged into the base model, and models without official list pricing as well as third-party modified versions are excluded.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/17/2026
First Seen
Valid until: 1/15/2027
Sources
Related Entities
Related Claims
UnverifiedDevelopers should pay attention to billing details (terms such as caching, batching, and concurrency limits) rather than headline prices, because the former may affect the bill more than the unit price.66% similarUnverified"Up to X% off" type discounts are usually calculated from the cheapest cell in the entire price catalog (such as the output price of a limited-time promotion for an unpopular model, or the first-purchase tier for new users), and are an upper bound rather than the expected value of a typical scenario.63% similarUnverifiedEvaluating document parsers requires locking the solution version number and recording per-page cost and latency alongside the pass rate, because a 2% higher pass rate that costs 10 times more is not necessarily worthwhile.63% similarUnverifiedThe formula for evaluating real API cost should be: nominal unit price × effective token count × (1 + retry rate), rather than only looking at the unit price on the pricing page.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1057022API
curl https://kongchang.com/api/v1/knowledge/claims/1057022MCP
get_claim(id=1057022)