OpenSearch LLM-Based Conversational Search Edition charges for storage and compute separately, with different billing models depending on the edition and billing method you choose.
Billing methods
| Billing method | Description |
|---|---|
| Pay-as-you-go | Allows immediate use of resources. Charges are calculated hourly and deducted from your Alibaba Cloud account balance. All hourly charges are consolidated under a single purchase order. |
| Subscription | Storage is paid upfront at instance creation, offering cost savings compared to pay-as-you-go. Computing resources for Standard Edition instances are still billed on a pay-as-you-go basis — purchase a computing resource plan to cover compute costs. |
Understand the billing model
Storage and compute are billed independently. For pay-as-you-go, both storage and compute are charged hourly based on actual usage. For subscription, you pay a fixed upfront fee for storage capacity, while compute continues to accrue hourly charges based on the number of conversational search requests your instance processes.
A subscription instance still generates variable compute charges each hour, even though storage is prepaid.
Billable items
| Edition | Billable item | Unit | Range | Step | Description |
|---|---|---|---|---|---|
| Standard Edition | Storage resources | GB | 1–100 | 1 GB | Fixed fee based on the storage capacity selected at purchase. |
| Standard Edition | Computing resources | CU (compute unit) | 0–3,600 | — | Charged hourly based on actual usage. Higher request volumes and queries per second (QPS) consume more CUs. |
| Professional Edition | Storage resources | GB | 1–100 | 1 GB | Fixed fee based on the storage capacity selected at purchase. |
| Professional Edition | GPU rentals | Unit | 1–10 | 1 | Fixed rental fee that varies by GPU specifications. |
Standard Edition instances share data processing resources with other instances. Latency may occur when uploading data.
Pricing — Standard Edition (Singapore)
Pay-as-you-go
| Billable item | Price | Example |
|---|---|---|
| Storage resources | USD 0.030/GB-hour | 10 GB × USD 0.030/GB-hour = USD 0.30 |
| Computing resources | USD 0.086/CU-hour | 100 CUs × USD 0.086/CU-hour = USD 8.60 |
Subscription
| Billable item | Price | Example |
|---|---|---|
| Storage resources | USD 15.12/GB-month | 2 GB × 1 month = USD 30.24 (paid upfront) |
Computing resources continue to be billed at the pay-as-you-go rate (USD 0.086/CU-hour).
Billing examples
Standard Edition — pay-as-you-go (Singapore, 2 GB)
| Time window | Requests | CUs consumed | Calculation | Total |
|---|---|---|---|---|
| 06:00–07:00 | 0 | 0 | 2 GB × USD 0.030 | USD 0.06 |
| 07:00–08:00 | 100 | 10 | 2 GB × USD 0.030 + 10 CUs × USD 0.086 | USD 0.92 |
| 08:00–09:00 | 1,000 | 100 | 2 GB × USD 0.030 + 100 CUs × USD 0.086 | USD 8.66 |
Standard Edition — subscription (Singapore, 2 GB, 2 months)
Storage (upfront): 2 GB × 2 months × USD 15.12/GB-month = USD 60.48
Compute (pay-as-you-go): charged hourly at USD 0.086/CU-hour based on actual usage
| Time window | Requests | CUs consumed | Compute charge |
|---|---|---|---|
| 06:00–07:00 | 0 | 0 | USD 0.00 |
| 07:00–08:00 | 100 | 10 | 10 CUs × USD 0.086 = USD 0.86 |
Usage notes
Compute unit (CU) consumption
On average, 1 CU handles approximately 10 conversational search requests. Actual consumption varies with conversation complexity.
Computing resources are billed based on actual usage, including Q&A calls made from both the console and the API.
Image content detection
If image content detection is enabled, images in uploaded files are automatically detected and billed as additional compute usage. Each image typically consumes 0.1–0.5 CUs.
QPS limit
Each instance supports a maximum of 10 queries per second (QPS) for conversational search. Exceeding this limit may affect the service level agreement (SLA). For higher QPS requirements, apply in advance.