Campaign Update Notice

Due to business adjustments, starting today, the “Flagship Model Special Subscription Package” and the “Lightweight All-in-One Subscription Package” will be officially merged into the “Special Resource Package,” with unified pricing adjustments. If you have previously purchased the above subscription products, your existing benefits will be fully retained and will not be affected. Please feel free to continue using them. If you have any questions about this adjustment, please feel free to contact customer service at any time.

Unified Token Description

The tokens in the subscription package are unified tokens and can be used to call any model included in the package. The specific consumption ratio is subject to real-time calculation during the API call.

• Benchmark definition: The input token price of the claude-haiku-4-5-20251001 model is used as the benchmark price, i.e., $1/Mtoken (1 USD per million input tokens).
• Conversion rule: When calling other models, the system will automatically convert and deduct unified tokens from the subscription package based on the ratio between that model’s listed price and the benchmark model price. For example, if a model’s input price is $3/Mtoken, the number of unified tokens consumed when calling that model will be 3 times that of the benchmark model.
• Consumption calculation: The token consumption for one API call is input token consumption + Cached writes token consumption + Cached reads token consumption + output token consumption, each calculated according to the respective model's Deduction Multiplier for conversion.

Example: In one inference request for claude-sonnet-4-6

• Token consumption: input tokens: 11, Cached writess (5m) tokens: 4629, output tokens: 2456
• Deduction multipliers: input: 3, Cached writess (5m): 3.75, output: 15
• Unified tokens consumed = 11 × 3 + 4629 × 3.75 + 2456 × 15 ≈ 54,231