---
type: "post"
id: "deepgemm-ascend-deepgemm-ascend-clean-and-efficient-matrix-ffbb59"
slug: "deepgemm-ascend-deepgemm-ascend-clean-and-efficient-matrix-ffbb59"
title: "DeepGEMM-Ascend: DeepGEMM-Ascend: clean and efficient matrix multiplication kernel library for Huawei Ascend…"
description: "Thesis: DeepGEMM-Ascend is a notable software port because it brings DeepSeek's matrix-kernel library to Huawei Ascend while retaining a familiar API."
retrieval_nugget: "Thesis: DeepGEMM-Ascend is a notable software port because it brings DeepSeek's matrix-kernel library to Huawei Ascend while retaining a familiar API."
published_at: "2026-09-30"
updated_at: "2026-09-30"
record_date: "2026-09-30"
date_kind: "published_at"
topics: []
source_urls: ["https://github.com/deepseek-ai/DeepGEMM-Ascend"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:deepgemm-ascend-deepgemm-ascend-clean-and-efficient-matrix-ffbb59"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:ef94ae989fb56201ea841d64879dced2783b53f91c00e6a57277d497e55ba6ec","reviewed_at":"2026-09-30T14:03:44.005Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":1,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"DeepGEMM-Ascend is a notable software port because it brings DeepSeek's matrix-kernel library to Huawei Ascend while retaining a familiar API.","observed_facts":[{"text":"The GitHub repository identifies DeepGEMM-Ascend as a matrix multiplication kernel library for Ascend NPUs and records an OSS release commit dated September 30, 2026.","source_urls":["https://github.com/deepseek-ai/DeepGEMM-Ascend"]},{"text":"The supplied release material lists FP4, FP8 and BF16 GEMM, grouped GEMM for MoE, MQA logits and fused MegaMoE operators.","source_urls":["https://github.com/deepseek-ai/DeepGEMM-Ascend"]},{"text":"The project reports dense GEMM performance of 1,701 TFLOPS FP4, 861 TFLOPS FP8 and 431 TFLOPS BF16.","source_urls":["https://github.com/deepseek-ai/DeepGEMM-Ascend"]}],"mechanism":"The project describes kernels rewritten for Ascend and reports using JIT compilation, coroutine pipelining and a thin wrapper over Ascend matrix instructions.","why_now":"DeepSeek has made the port publicly available as an OSS release, creating a concrete artifact developers can inspect and test rather than a roadmap claim.","implications":["The port could reduce the software friction of moving DeepSeek-style matrix workloads onto Ascend hardware, particularly for MoE-oriented operations."],"evidence_boundary":"The source establishes the repository, listed capabilities, requirements and project claims; it does not independently validate benchmarks or establish production adoption.","watch_conditions":["Whether independent users can reproduce the reported utilization on specified Ascend hardware and software versions."],"related_records":[],"new_branch_reason":"No prior New Runtime record was supplied, so this public software release starts a distinct analytical record rather than being attached to an unverified related item."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/deepgemm-ascend-deepgemm-ascend-clean-and-efficient-matrix-ffbb59/","markdown":"https://newruntime.com/posts/deepgemm-ascend-deepgemm-ascend-clean-and-efficient-matrix-ffbb59.md","json":"https://newruntime.com/posts/deepgemm-ascend-deepgemm-ascend-clean-and-efficient-matrix-ffbb59.json"}
---

# DeepGEMM-Ascend: DeepGEMM-Ascend: clean and efficient matrix multiplication kernel library for Huawei Ascend…

## Retrieval answer

Thesis: DeepGEMM-Ascend is a notable software port because it brings DeepSeek's matrix-kernel library to Huawei Ascend while retaining a familiar API.

Thesis: DeepGEMM-Ascend is a notable software port because it brings DeepSeek's matrix-kernel library to Huawei Ascend while retaining a familiar API. The GitHub repository identifies itself as a matrix multiplication kernel library for Ascend NPUs, and its release material lists FP4, FP8 and BF16 GEMM support alongside grouped GEMM for MoE, MQA logits and fused MegaMoE operators.

Mechanism: The project describes kernels rewritten for Ascend and reports using JIT compilation, coroutine pipelining and a thin wrapper over Ascend matrix instructions. Why now: DeepSeek has made the port publicly available as an OSS release, creating a concrete artifact developers can inspect and test rather than a roadmap claim. The repository commit is dated September 30, 2026. Its reported dense GEMM figures are 1,701 TFLOPS for FP4, 861 TFLOPS for FP8 and 431 TFLOPS for BF16, corresponding to 98.3%, 99.5% and 99.8% of theoretical peak.

Implication: The port could reduce the software friction of moving DeepSeek-style matrix workloads onto Ascend hardware, particularly for MoE-oriented operations. What to watch: Whether independent users can reproduce the reported utilization on specified Ascend hardware and software versions. Evidence boundary: The source establishes the repository, listed capabilities, requirements and project claims; it does not independently validate benchmarks or establish production adoption. The documented requirements include CANN 9.20, torch_npu and Python 3.10+, and the repository specifies an MIT license.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)