DeepSeek Open-Sources Software Tools for Huawei AI Chips
Source: The Information (opens in a new tab) · Juro Osawa
Intel Summary
DeepSeek announced plans to open-source software tools customized for Huawei AI chips, including a compatible version of TileLang. The initiative supports the company's efforts to train and run its artificial intelligence models on domestic semiconductors.
Why It Matters
Software stack compatibility is a primary bottleneck for adopting alternative AI accelerators; open-sourcing tooling for Huawei silicon lowers the barrier for developers and enterprises seeking to deploy models on non-Nvidia hardware.
Part of an ongoing development
SourceDeepSeek releases TileLang open-source programming software for Huawei Ascend chips
DeepSeek and Huawei have released open-source programming tools for Huawei's Ascend AI hardware, centered around a programming language called TileLang. The tooling is designed to offer a simpler programming model than Nvidia's CUDA and optimize software performance on domestic Chinese AI chips. Claims are as reported; this summary makes no determination about accuracy or significance.
More coverage of this development
Organizations & Entities
Topics
Related Intelligence
- ReportSame development
China's AI industry closes ranks as Deepseek ships open-source software for Huawei's Ascend chips
DeepSeek and Huawei have released open-source programming tools for Huawei's Ascend AI hardware, centered around a programming language called TileLang. The tooling is designed to offer a simpler programming model than Nvidia's CUDA and optimize software performance on domestic Chinese AI chips.
The Decoder - DevelopmentStableAlso involving DeepSeek
Anthropic details distillation campaigns from Alibaba, Moonshot AI, and DeepSeek
TechCrunch reports that Anthropic has published a report alleging persistent model distillation campaigns conducted by China-based AI companies, specifically citing Alibaba, Moonshot AI, and DeepSeek. Claims are as reported; this summary makes no determination about accuracy or significance.
2 independent sources - DevelopmentStableAlso involving DeepSeek
DeepSeek released V4.1-Flash
DeepSeek has released V4.1-Flash, a multimodal model containing 552 billion total parameters with 16 billion active parameters per token. According to benchmark results reported from DeepSeek, V4.1-Flash narrowly outperforms Opus 5 and GPT-5.6 Sol on the DeepSWE coding benchmark and is published under an open MIT license. Claims are as reported; this summary makes no determination about accuracy or significance.
2 independent sources - DevelopmentNewAlso involving DeepSeek
DeepSeek reaches $1 billion annualized revenue and finalizes $7.5 billion fundraising
The Information reports that DeepSeek's annualized revenue run rate has reached $1 billion, rising from less than $500 million several months ago due to price increases and ongoing model adoption. According to sources with direct knowledge, CEO Liang Wenfeng shared the figure with investors as the startup works to close a $7.5 billion financing round by late October ahead of a planned Shanghai Stock Exchange listing. Claims are as reported; this summary makes no determination about accuracy or significance.