This topic describes the billing for OpenSearch-LLM AI Chat Edition.
Billing methods
Billing method | Description |
Pay-as-you-go | This is a post-paid method. A bill is generated for your usage each hour. All hourly bills are part of a single order, and the fees are deducted from your Alibaba Cloud account. |
Subscription | This is a pre-paid method. Pay for the instance when you create it. This method is more cost-effective than pay-as-you-go. For Standard Edition instances, computing resources are still billed on a pay-as-you-go basis. Use a computing resource package for better value. |
Computing resource package | A computing resource package is a discount plan for the AI Chat Edition. Pre-pay for a package that contains a set amount of computing resources. This lets you get computing resources at a lower price. |
Billable items
LLM edition | Billable item | Unit | Purchase range | Step size | Description |
Standard Edition | Storage | GB | [1, 100] | 1 GB | Fixed storage fee. You are billed based on the quota you select at purchase. |
Computing resource | CU | [0, 3600] | - | Billed based on actual hourly usage. Higher call volumes and QPS result in greater consumption of computing resources. Note Importing data that contains images incurs image detection fees. These fees are included in the computing resource costs. | |
Professional Edition | Storage | GB | [1, 100] | 1 GB | Fixed storage fee. You are billed based on the quota you select at purchase. |
GPU rental | Unit | [1, 10] | 1 | Fixed rental fee. You are billed based on the unit price of the selected GPU specification. |
The data processing resources for Standard Edition instances are shared with other instances. You may experience delays when you upload data.
Billing rules
Standard Edition instances
Billing method: Pay-as-you-go
Billable item
Region
Price (CNY/GB/hour)
Billing example
Storage
Shanghai
0.232
Instance A is purchased with a storage quota of 10 GB. The hourly fee is: 10 × 0.232 = CNY 2.32.
Singapore
0.207
Instance A is purchased with a storage quota of 10 GB. The hourly fee is: 10 × 0.207 = CNY 2.07.
Billable item
Region
Price (CNY/CU/hour)
Billing example
CU computing resource
Shanghai
0.53
Instance A uses 100 CUs of computing resources for search and Q&A from 11:00 to 12:00. The fee for that hour is: 100 × 0.53 = CNY 53.
Singapore
0.592
Instance A uses 100 CUs of computing resources for search and Q&A from 11:00 to 12:00. The fee for that hour is: 100 × 0.592 = CNY 59.2.
Billing method: Subscription
Billable item
Region
Price (CNY/GB/month)
Billing example
Storage
Shanghai
116.91
You purchase an instance with 2 GB of storage. On the order confirmation page, you must pay CNY 223.82. (Computing resources are billed separately on a pay-as-you-go basis.)
Singapore
104.1
You purchase an instance with 2 GB of storage. On the order confirmation page, you must pay CNY 208.2. (Computing resources are billed separately on a pay-as-you-go basis.)
Billing method: Computing resource package
Billable item
Region
Computing resource package specification (CU)
Price (CNY)
CU computing resource
Shanghai
100
52.5
1000
519
10000
4770
Singapore
100
58.6
1000
580
10000
5330
Professional Edition instances
Professional Edition instances are not currently available for purchase in the Singapore region.
Billing method: Pay-as-you-go
Billable item
Region
Price (CNY/GB/hour)
Billing example
Storage
Shanghai
0.232
Instance A in the Shanghai region is purchased with a storage quota of 10 GB. The hourly fee is: 10 × 0.232 = CNY 2.32.
Billable item
Region
GPU specification
Price (CNY/unit/hour)
Billing example
GPU rental
Shanghai
gpu.v100.32g.x4
5.633
Instance A in the Shanghai region is purchased with two `gpu.v100.32g.x4` GPUs. The hourly fee is: 5.633 × 2 = CNY 11.266.
gpu.v100.32g.x8
11.263
Billing method: Subscription
Billable item
Region
Price (CNY/GB/month)
Billing example
Storage
Shanghai
116.91
You purchase an instance with 2 GB of storage. On the order confirmation page, you must pay CNY 223.82.
Billable item
Region
GPU specification
Price (CNY/unit/month)
Billing example
GPU rental
Shanghai
gpu.v100.32g.x4
2839.3
You purchase one `gpu.v100.32g.x4` GPU in the Shanghai region. On the order confirmation page, you must pay CNY 2839.3.
gpu.v100.32g.x8
5676.3
Computing resources are billed based on actual usage. This includes resources consumed by Q&A calls from the console and APIs, and by image detection during data import.
On average, 1 CU of computing resources can support 10 Q&A calls. Actual consumption depends on the complexity of the conversation.
The maximum access limit for each instance is 10 QPS. The Service-Level Agreement (SLA) is not guaranteed for traffic that exceeds 10 QPS. To increase the access limit, you can submit a request in advance.
After you purchase GPUs for a Professional Edition instance, you do not need to pay for computing resources.
GPU specifications cannot be changed, and downgrades are not supported. To change specifications or reduce the number of GPUs, you must purchase a new instance.
Custom LLMs
Billable item | Region | Model training price (CNY/billable hour) | Estimated billable hours for model training |
Computing resource CU | Shanghai | 3.87 | Training with the minimum dataset consumes 250 billable hours. To add more training resources, contact technical support. |
Computing resource packages can only be used to offset the cost of computing resources for search and Q&A. These packages cannot be used for model training. Custom LLMs only support the pay-as-you-go billing method. You are billed for the actual computing resources consumed during training.
Billing examples
Standard Edition instances
Billing method | Scenario | Billing example |
Pay-as-you-go | User A purchases a pay-as-you-go Standard Edition instance of OpenSearch-LLM AI Chat Edition. The user selects 2 GB of storage capacity in the Shanghai region. |
|
Subscription | User A purchases a subscription Standard Edition instance of OpenSearch-LLM AI Chat Edition. The user selects 2 GB of storage capacity for a 2-month subscription duration in the Shanghai region. |
|
Computing resource package | User A purchases a computing resource package for OpenSearch-LLM AI Chat Edition with a specification of 1,000 CUs. |
|
Professional Edition instances
Billing method | Scenario | Billing example |
Pay-as-you-go | User A purchases a pay-as-you-go Professional Edition instance of OpenSearch-LLM AI Chat Edition. The user selects 2 GB of storage capacity and two `gpu.v100.32g.x4` GPUs in the Shanghai region. | The hourly fee for this instance is: 2 × 0.232 + 2 × 5.633 = CNY 11.73. |
Subscription | User A purchases a subscription Professional Edition instance of OpenSearch-LLM AI Chat Edition. The user selects 2 GB of storage capacity and two `gpu.v100.32g.x4` GPUs in the Shanghai region. | On the order confirmation page, the user must pay: 2 × 116.91 + 2 × 2839.3 = CNY 5912.4. |
Custom LLMs
Scenario | Billing example |
User A purchases a Professional Edition instance (subscription or pay-as-you-go) of OpenSearch-LLM AI Chat Edition and chooses to train a custom LLM. | Custom LLM training typically takes 2 to 3 business days to complete. It consumes a minimum of 250 CUs of training resources. The fee for the custom LLM is: 250 × 3.87 = CNY 967.5. |