Developers can now use DeepSeek's open math code on Huawei Ascend chips
DeepSeek released DeepGEMM-Ascend on September 30, a free matrix-math library for Huawei Ascend chips. Reports say it is one of six modules in the release.
DeepSeek shared free computer code that helps AI programs run on Huawei's AI chips. Anyone can read and use it. It is mostly useful for people who build AI software.
Hype check
The noise matches the real change.
Our editorial opinion, based on the sources below. How we score
Your call: hype or real?
Real = it will matter in daily use. Hype = louder than it is. One vote, and it can’t be changed.
Our view and readers’ views are opinions, not statements of fact about a product.
Coming your way?
Available worldwide · Free (MIT license)
The code is free for anyone to download. It is built for Huawei Ascend chips, so you need that hardware to run it.
Why it matters
Open tools make it easier to run AI on more kinds of chips. It matters most to developers who work with Ascend hardware.
Watch: more open tools help AI run on more kinds of chips.
The details
- DeepSeek's GitHub lists DeepGEMM-Ascend under the MIT license, first released on September 30, 2026.
- The library handles matrix math (GEMM) in BF16, FP8 and FP4 on Huawei Ascend chips.
- It was developed and validated on Ascend 950 series devices.
- The README says it reaches up to 99.8% hardware use. That is DeepSeek's own figure.
- The Next Web reports six modules in all, including compute and communication libraries and Ascend support for TileLang, with Huawei's help.
Receipts
The company or lab said it. Primary source linked. Open the sources and check us.
Written with the help of AI tools and checked against the sources above. How we work
See a mistake? Tell us and we’ll fix it.
