Token Plan
A prepaid subscription model offered by large model service providers. Users purchase a fixed token quota in advance, which is deducted based on actual API call consumption within the validity period. The plan typically enforces dual constraints: a total token cap and an expiration date.
Timeline (last 90 days)
Alibaba Cloud Bailian has launched a Token Plan subscription service focused on monthly bundled model-call quotas.
A Bilibili UP host subscribed to a mid-tier Bailian Token Plan priced at 139 yuan.
From 11:45 PM that evening to 3:09 AM, over a span of more than three hours, the UP host exhausted the plan's entire seven-day quota limit.
The total consumption in this instance reached 124 million (1.24 × 10⁸) tokens, and this occurred with a cache hit rate as high as 99%.
When the UP host asked a Bailian engineer about the consumption ratio among different models in the Token Plan, the engineer was unable to provide a clear answer.
Token Plans with shared quotas across multiple models involve model conversion rate issues, where the billed tokens consumed per call for different models may not equal the actual number of text tokens processed.
For applications with high-frequency calls, usage caps and alerts should be set to prevent quotas from being exhausted unattended.
All Facts (3)
Token Plans with shared quotas across multiple models involve model conversion rate issues, where the billed tokens consumed per call for different models may not equal the actual number of text tokens processed.
50%UnverifiedFor applications with high-frequency calls, usage caps and alerts should be set to prevent quotas from being exhausted unattended.
50%UnverifiedAlibaba Cloud Bailian has launched a Token Plan subscription service focused on monthly bundled model-call quotas.
50%