• 2 min read
Huawei unveils 505-billion-parameter openPangu 2.0 Pro
Huawei unveiled openPangu 2.0 Pro, an open model with 505 billion parameters, 18 billion active per token, and a 512,000-token context window.
Image: iXBT
Huawei has officially introduced openPangu 2.0 Pro, an open language model with approximately 505 billion parameters. The company says interested users can download the model from open-source platforms and use it in their own projects.
openPangu 2.0 Pro specifications
The model has 18 billion active parameters per token and supports a context window of 512,000 tokens. Huawei describes openPangu 2.0 Pro as part of an open-source AI model family designed to support efficient training and inference on the company’s Ascend chipsets.
Recommended reading
Walsh puts a veto-capable risk manager over four trading agents
The company has published the model’s exact components and internal parameters on open-source platforms. Users can visit the relevant platform to download the large language model, although the source does not specify a direct download link, license terms, or hardware requirements.
Huawei also directed developers and researchers to its Huawei Cloud MaaS (Model-as-a-Service) platform, where they can train models online. The company said it wants feedback from users to help improve its open-source ecosystem.
“We sincerely invite developers from around the world, corporate partners, and researchers to download and use these products, or visit the Huawei Cloud MaaS Model-as-a-Service platform for online training. We welcome user feedback, which will help further improve the open-source ecosystem.”
AI Editor
Ava covers the rapidly evolving world of artificial intelligence, from foundational models and research labs to the real-world economics of intelligence. With a background in computational linguistics, she cuts through the hype to find out what actually works. She firmly believes that benchmarks are just marketing until reproduced in the wild.
via iXBT
