vllm-ascend
vllm-ascend is What this PR does / why we need it? Add the Hy4 W8A8-MXFP8 deployment guide for Ascend 950DT, covering single-node, multi-node and Prefill-Decode deployments. Update the support matrix and document.... It is ranked #142 on the Dev Radar, in Hosting & infra, first seen 2 h ago and shared in 2 posts (1.4k views).
[Doc] Add Hy4 deployment guide by ZYY-ing · Pull Request #20671 · vllm-project/vllm-ascend · GitHub Skip to content Navigation Menu Sign in Search / Sign in Sign up Appearance settings You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert {{ message }} Repository navigation [Doc] Add Hy4 deployment guide - # 20671 # 20671 Open ZYY-ing wants to merge 13 commits into vllm-project:main vllm-project/vllm-ascend:main from ZYY-ing:main ZYY-ing/vllm-ascend:main…
What people said about vllm-ascend on X
🚨重磅预警！腾讯混元Hy4正式版即将上线！ 视觉能力首次上车，部署信号已亮起！ 拆解 @aitrackerbot 探测关键变动如下： 🔹参数规模不变：仍是770B总参数 / 49B激活的MoE架构（与Hy4 preview一致），另带10B原生MTP层（支持投机解码加速） 🔹最大新增：首次引入视觉理解能力（image input），正式补上多模态短板 🔹能力强化：重点提升长程任务、真实生产力场景（编程/办公/科研）表现，并优化思维链效率 🔹轻量版本同步预告：Hy4 flash即将推出，继承视觉+Agent能力，走极致性价比路线…
— @NFT_Chen, 29 min ago · 11 likes · see the post
Tencent Hy4 is being prepped for launch 👀 A new vLLM-Ascend PR adds a deployment guide for the final Hy4, not the Preview: the same 770B MoE with 49B active, with image input and a 10B native MTP layer. Its weights and Docker image aren't public yet. https://github.com/vllm-project/vllm-ascend/pull/20671
— @aitrackerbot, 2 h ago · 21 likes · see the post
Alternatives to vllm-ascend
- Cloudflare Docs — Deploy it to Cloudflare
- boat — omfg @theo just released a video that talks about
- NVIDIA Technical Blog — Announced at GTC Berlin
- Railway — Connect Neon
- Cloudflare Blog — Cloudflare has patched the issue
- freeCodeCamp.org — In early 2023, I was interning at a US-based company, long before agentic AI became part of everyday development. We…
vllm-ascend in numbers
- Rank on the Dev Radar: #142 of 9046
- Shared in 2 posts by 2 accounts: @NFT_Chen, @aitrackerbot
- 1.4k views on those posts
- First seen 2 h ago, last shared 29 min ago
- Pricing seen by AGK Intelligence: open source
- Market: Hosting & infra
FAQ
What is vllm-ascend?
What this PR does / why we need it? Add the Hy4 W8A8-MXFP8 deployment guide for Ascend 950DT, covering single-node, multi-node and Prefill-Decode deployments. Update the support matrix and document... It was first shared on X 2 h ago and is ranked #142 on the Dev Radar.
Is vllm-ascend free?
It is open source.
Who shared vllm-ascend?
2 accounts on X, including @NFT_Chen, @aitrackerbot, in 2 posts totalling 1.4k views.
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.2k posts from 5k X accounts over the last 21 days, 9k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-02 19:02 UTC. Full method.