To further enhance the service capabilities of our university's AIGC platform and better support faculty and students in utilizing large language models for teaching, research, and daily office work, the Information Technology and Data Intelligence Department plans to perform a service upgrade on the AIGC platform. The relevant details are hereby notified as follows:
I. Upgrade Schedule
September 4, 2026 (Friday)18:00 – 22:00
II. Service Impact
During the upgrade, the following services will be temporarily unavailable:
• Chat Interface Service:https://aigc.hkust-gz.edu.cn
• API Service: https://aigc-api.hkust-gz.edu.cn/v1/*
III. Upgrade Details
1. Chat Interface Service Upgrade
|
Model Name |
BackendModel Version |
Free Allowance |
Status |
|
GPT-5.6 |
gpt-5.6-luna |
500,000 tokens, suspended upon exceeding the limit |
Newly Added |
|
GLM-5.2 |
GLM-5.2 |
Unlimited |
Newly Added |
|
DeepSeek-V4-Flash |
DeepSeek-V4-Flash |
Unlimited |
Retained |
|
DeepSeek-V4-Pro |
DeepSeek-V4-Pro |
Unlimited |
Retained |
|
ChatGPT |
gpt-4o-mini |
/ |
Retired |
|
GPT-4 |
gpt-4o |
/ |
Retired |
|
GPT-4 Vision |
gpt-4o |
/ |
Retired |
2. API Service Upgrade
(1) API Endpoint Updates:
• V2 API Endpoint: The V2 Endpoint supports frontend parameter passthrough, allowing call parameters to be passed directly to the backend large model gateway
• V1 API Endpoint: After this upgrade, all calls to V1 endpoints will be automatically forwarded to V2 endpoints for processing
• To avoid compatibility risks, it is recommended that existing users / systems begin using V2 endpoints for business testing and verification from now on
• API reference guide: https://docs.hpc.hkust-gz.edu.cn/docs/aigc/api-usage
|
API Version |
API Endpoint |
|
V2 |
|
|
V1 |
|
(2) Model Updates
The new models GPT-5.6, GLM-5.2, Kimi-K3, and Qwen-3.6 are now available. The complete list of models supported by the AIGC platform and their pricing are shown below. We kindly remind you that on-campus computing resources are limited. Please manage request concurrency to avoid overloading backend resources. API calls are billable by default with no free quota. We recommend prioritizing on-campus local models and planning your budget accordingly to use the service responsibly and prevent overages.
In the future, the platform will be further optimized in areas such as free quota, key management, usage monitoring, concurrency control, and cache-based billing.
|
Model Type |
Model Name (API call parameter value) |
Backend Model Version |
Model Source |
Prompt Pricing (per 1M) |
Completion Pricing (per 1M) |
This Update |
|
Chat |
gpt-5.6-luna |
gpt-5.6-luna |
External Service |
1.4662 |
8.7996 |
Newly Added |
|
Chat |
gpt-5.6-terra |
gpt-5.6-terra |
External Service |
14.662 |
87.996 |
Newly Added |
|
Chat |
gpt-4o |
gpt-4o-2024-08-06 |
External Service |
18.328 |
73.310 |
Newly Added |
|
Chat |
gpt-4o-mini |
gpt-4o-mini-2024-07-18 |
External Service |
1.110 |
4.399 |
Newly Added |
|
Embedding |
text-embedding-3-large |
text-embedding-3-large |
External Service |
0.95310 |
/ |
Price Updated |
|
Embedding |
text-embedding-3-small |
text-embedding-3-small |
External Service |
0.14663 |
/ |
Price Updated |
|
Chat |
GLM-5.2 |
GLM-5.2 |
On-Campus Local |
4.00 |
14.00 |
Newly Added |
|
Chat |
Kimi-K3 |
Kimi-K3 |
On-Campus Local |
10.00 |
50.00 |
Newly Added |
|
Chat |
Qwen3.6-35B-A3B |
Qwen3.6-35B-A3B |
On-Campus Local |
0.90 |
5.40 |
Newly Added |
|
Chat |
DeepSeek-V4-Pro |
DeepSeek-V4-Pro |
On-Campus Local |
2.25 |
6.75 |
Price Updated |
|
Chat |
DeepSeek-V4-Flash |
DeepSeek-V4-Flash |
On-Campus Local |
0.75 |
2.25 |
Price Updated |
(3) Model Retirement: The following models will be officially retired during the upgrade on September 4, 2026. Please switch to the latest models in advance.
|
Model Name (API call parameter value) |
Backend Model Version |
|
gpt-5.3-chat |
gpt-chat-latest |
|
Qwen3.0 |
Qwen3-235B |
|
gpt-4o-2024-08-06 |
gpt-4o-2024-08-06 |
|
gpt-4-turbo-app |
gpt-4o-2024-02-01 |
|
gpt-4-turbo |
gpt-4o-2024-08-06 |
|
gpt-3.5-turbo-app |
gpt-4o-mini-2024-07-18 |
|
text-embedding-ada-002 |
text-embedding-ada-002 |
IV. AIGC Service Application and Usage
(1) The Chat Interface Service is accessible by default to all staff and students at: https://aigc.hkust-gz.edu.cn.
(2) The API Service requires an application submitted by staff through DingTalk. Application process: Staff members submit an "AIGC Service Application" via the DingTalk OA system, selecting "API Mode" as the application type. Upon approval, the administrator will issue an API management account.
(3) API Keys required for invoking large language models must be created by the staff members, and may be assigned to RAs or students within their research group for use. The staff member bears all related management responsibilities. All costs incurred from API calls will be charged to the staff member's account and the bills will be sent to their PBMS account at the beginning of the following month.
(4) For instructions on API Key creation and call operations, please refer to the HPC Knowledge Base: https://docs.hpc.hkust-gz.edu.cn/docs/aigc/api-key
If you have any comments or suggestions regarding this upgrade, please feel free to contact the HPC team: hpc@hkust-gz.edu.cn
Information Technology and Data Intelligence Department (ITDID)
September 2, 2026
