AnswersPublished18 sources
Huawei Cloud’s openPangu-2.0 Retirement: Key Dates and What Customers Should Do
Huawei Cloud will stop selling both openPangu 2.0 MaaS services at 00:00 Beijing time on November 6, 2026 , then retire them at 00:00 on November 30 . Huawei recommends DeepSeek V4.1 Flash 1017 for Flash and GLM 5.3 for Pro; test API compatibility, pricing and performance rather than assuming either is a drop in rep...
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- Huawei Cloud will stop selling both openPangu 2.0 MaaS services at 00:00 Beijing time on November 6, 2026 , then retire them at 00:00 on November 30 .
- Huawei recommends DeepSeek V4.1 Flash 1017 for Flash and GLM 5.3 for Pro; test API compatibility, pricing and performance rather than assuming either is a drop in replacement.[43]
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What does Huawei Cloud’s November 2026 retirement notice mean for openPangu-2.0-Flash and openPangu-2.0-Pro customers on its Model as a Serv. Article summary: Huawei Cloud is retiring the hosted MaaS services for openPangu-2.0-Flash and openPangu-2.0-Pro: sales stop at **00:00 Beijing time on November 6, 2026**, and shutdown is scheduled for **00:00 Beijing time on November 30. Topic tags: general, general web, documentation. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fak
Huawei Cloud is retiring the hosted MaaS services for openPangu-2.0-Flash and openPangu-2.0-Pro. Sales stop at 00:00 Beijing time (UTC+8) on November 6, 2026. Customers who have already purchased a service can continue using it during the transition, but both models are scheduled to go offline at 00:00 Beijing time on November 30. After that, neither new nor existing users can call the hosted services.2
43
The practical deadline is midnight at the start of November 30—not the end of that day. Plan and test a replacement before then.
Retirement dates and affected versions
Huawei Cloud’s MaaS model listing shows openPangu-2.0-Flash version 20260630 and openPangu-2.0-Pro version 20260730 in the Southwest-Guiyang1 region.6 Huawei’s OpenAI-compatible interface documentation also lists both models for that region.
19 Customers using these hosted versions should confirm their own deployment and region in the MaaS console before scheduling a cutover.
| Milestone | Beijing time | What it means |
|---|---|---|
| Sales stop | November 6, 2026, 00:00 | New users can no longer purchase or activate either service; existing customers can continue during the transition. |
| Services go offline | November 30, 2026, 00:00 | New and existing users can no longer call either hosted model. |
Recommended replacements
Huawei’s reported recommendations are DeepSeek-V4.1-Flash-1017 for openPangu-2.0-Flash and GLM-5.3 for openPangu-2.0-Pro.43 These are alternatives to evaluate, not evidence that the replacements behave identically to the retiring models.
Before switching, compare the models against your production needs:
- Capabilities: Check context and output limits, reasoning controls, tool or function calling, structured responses and any multimodal needs. Huawei’s MaaS updates list 1M context and function calling for both recommended models, but that alone does not establish equivalent task performance.
50
- API behavior: Compare supported request parameters, defaults, tool schemas, streaming events, response fields and error handling. Validate any model-specific reasoning or caching options against the current documentation.
- Pricing and limits: Check current rates, token accounting, quotas and concurrency limits in MaaS. The available sources do not provide enough information for a reliable price comparison.
- Real workload results: Run representative prompts and compare correctness, tool-use reliability, latency, throughput and cost per completed task. A replacement recommendation is not a performance guarantee.
43
Migration checklist for MaaS APIs
The supplied retirement coverage does not specify separate migration instructions for MaaS standard API V1, standard API V2 or OpenAI-compatible interfaces. In particular, it does not establish that changing only the model name is sufficient across those interfaces.
Huawei’s OpenAI-compatible API documentation identifies an endpoint and model IDs for the current openPangu services, but that is not a retirement-specific guide or confirmation that the recommended replacements use identical parameters and response behavior.19 Before cutover:
- Confirm that the replacement is available in your region and supports the interface you use.
- Verify its endpoint, authentication method, model ID and request/response schema in current documentation.
- Review model-specific parameters and update request construction or response parsing where needed.
- Test streaming, tool calls, retries, error handling, quotas and billing.
- Run the new model against production-like tasks and keep a tested fallback until migration is complete.
Hosted service retirement is not the same as model-file removal
The notice concerns Huawei Cloud’s hosted MaaS endpoints. Separately, Huawei announced the open-source release of openPangu-2.0-Pro weights, basic inference code and a technical report on July 31, 2026.5 A model repository for openPangu-2.0-Flash is also listed on Hugging Face.
11 The end of the hosted service does not, by itself, establish that released model files are being withdrawn. Nor does their separate availability guarantee continued hosting, maintenance, licensing terms or easy self-hosting.
For context, Huawei’s model documentation describes Flash as an Ascend NPU-trained MoE model with about 92 billion total parameters, 6 billion activated per token and a 512K context window. It describes Pro as an Ascend NPU-trained MoE model with about 505 billion total parameters, 18 billion activated per token and a 512K context window.12 Those model specifications do not ensure that a self-hosted setup will match the managed MaaS service.
Don’t rely on an undated successor
Huawei has reportedly said that new openPangu versions are planned for gradual release, but specific model names and launch dates had not been disclosed in the report available on November 3, 2026.43 There is no confirmed successor timetable to rely on for continuity before the November 30 shutdown.
If a workload is waiting for a new openPangu release, use a tested interim replacement or assess self-hosting as a separate project. Keeping the model identifier and provider configurable can also make future changes easier to test without a broad application rewrite.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "Huawei Cloud’s openPangu-2.0 Retirement: Key Dates and What Customers Should Do"?
Huawei Cloud will stop selling both openPangu 2.0 MaaS services at 00:00 Beijing time on November 6, 2026 , then retire them at 00:00 on November 30 .
What are the key points to validate first?
Huawei Cloud will stop selling both openPangu 2.0 MaaS services at 00:00 Beijing time on November 6, 2026 , then retire them at 00:00 on November 30 . Huawei recommends DeepSeek V4.1 Flash 1017 for Flash and GLM 5.3 for Pro; test API compatibility, pricing and performance rather than assuming either is a drop in replacement.[43]