Open-source Pangu Model openPangu-2.1-Pro Launches on Huawei Cloud, with Capability Upgrades Around Agent Development, AI Coding, and Inference Optimization
IT Home, November 15 news: Huawei Cloud announced today that the Huawei Cloud MaaS (Model-as-a-Service) platform has officially launched the open-source Pangu model openPangu-2.1-Pro.
The model upgrades its capabilities around three major directions—Agent development, AI coding, and inference optimization—and achieves enhancements in scenarios such as deep research, HarmonyOS system operation, HarmonyOS App development, Ascend operator generation, and inference pipeline optimization.

- In terms of Agent capabilities, openPangu-2.1-Pro optimizes Skills triggering, tool selection, and invocation capabilities based on more than 100,000 specialized data entries, improving complex task planning and execution results, and achieving capability enhancements in scenarios such as deep research and HarmonyOS system operation.
- In terms of coding capabilities, the new model focuses on optimization and adaptation for HarmonyOS App development and Ascend operator generation. Around the HarmonyOS ArkTS language and related development tasks, it continuously improves the accuracy of code generation and application usability; it carries out targeted optimization for the Ascend operator development process, enhances professional code generation capabilities, and provides support for developers to lower the development threshold and improve R&D efficiency.
- Through optimization of key operators and the inference pipeline, openPangu-2.1-Pro achieves a substantial increase in throughput efficiency in typical deployment scenarios, supporting high-concurrency applications with higher resource utilization efficiency. At the same time, through the preset model API, invocation rate limiting, usage statistics, and performance monitoring capabilities provided by Huawei Cloud MaaS, users can achieve rapid access and operations.
openPangu-2.1-Pro will also be integrated into core products such as Huawei Cloud Zhiguo AgentArts, MaDao CodeArts, and Guoban OfficeAce, covering agent development, AI coding, and enterprise office work, accelerating the extension of model capabilities into actual business scenarios.
- Inference input: 0.8 yuan / million tokens
- Inference output: 3.2 yuan / million tokens
- Cache hit input: 0.016 yuan / million tokens
Related Articles
- The first Ascend-native world model of the "Industry AI Dream Factory," Huawei Cloud CloudRobo Embodied World Model launch event scheduled for November 16
- Just one step away from the 100 million milestone, Huawei HarmonyOS installations reported to have exceeded 99 million units
- Dongfeng Honda CR-V All-Around Glory Edition new car launched: Huawei Cloud technology empowers cloud in-car applications, current price starting at 137,900 yuan
- Another dream combination realized: netizen successfully installs Huawei HarmonyOS 6 system on Apple MacBook Pro
- Starting at 6,899 yuan, Huawei Qingyun HM650/g commercial office HarmonyOS desktop computers go on sale December 1
- Huawei Qingyun HM650 / HM650g is aimed at government and enterprise customers and confirmed not to be sold in e-commerce or offline experience stores