Unverified50% confidenceFactExact time
Zhipu changed its quota algorithm from billing by number of uses to billing by credits: Lite gets 10,000 credits per week and 2,000 credits per 5 hours; Pro is 6 times Lite, and Max is 14 times.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/27/2026
First Seen
Valid until: 1/25/2027
Sources
Domestic AI Coding Plans Compared: How to Choose Between Zhipu, Kimi, and MiniMax
bilibili优哥跨境10/17/2026
Related Entities
Related Claims
Unverified顶级模型每次调用的算力消耗可能是轻量模型的数十倍68% similarUnverifiedVectorized execution processes data in batches (usually hundreds to thousands of rows), amortizing the interpretation overhead per row.63% similarUnverifiedThe turbo version delivers 20x faster output but at 10 times the price, reflecting the engineering trade-off between latency and throughput in inference compute, as acceleration requires allocating more parallel compute resources per request.61% similarUnverifiedA third-party test classifying 1,000 emails found the judgment model took 15.6 seconds at a cost of 1.77, while the large model took 48.9 seconds at a cost of 2.07—approximately 15% lower cost but roughly three times faster.61% similarUnverifiedOn ARM Cortex-M-class hardware, with the buoy performing 288 inferences per day, the naive per-sample update scheme consumes about 3686 millijoules/day, while updating only about 8% of the time with gating consumes only 564 millijoules/day, a 6.5-fold difference.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1101936API
curl https://kongchang.com/api/v1/knowledge/claims/1101936MCP
get_claim(id=1101936)