AnswersPublished15 sources
GPUWorks: A Lightweight, Open-Source Platform for AI Inference
ZStack released GPUWorks under Apache 2.0 on November 4, 2026. The roughly 40MB platform is designed to manage mixed accelerator clusters without Kubernetes; its reported scale benchmarks come from the company, not in...
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- ZStack released GPUWorks under Apache 2.0 on November 4, 2026. The roughly 40MB platform is designed to manage mixed accelerator clusters without Kubernetes; its reported scale benchmarks come from the company, not in...
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What is GPUWorks, the enterprise AI inference platform open-sourced on November 4, 2026, according to IT Home, and how does it combine Apach. Article summary: GPUWorks is ZStack’s lightweight, enterprise-grade AI inference and Model-as-a-Service (MaaS) platform, announced as open source under Apache-2.0 on November 4, 2026, with a public `flitai-io/gpuworks` GitHub repository.. Topic tags: general, general web. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake numbers, clic
GPUWorks is ZStack’s lightweight GPU-management and Model-as-a-Service platform for deploying AI inference across accelerator infrastructure. ZStack announced it as open source under the Apache-2.0 license on November 4, 2026, and the project is available in the flitai-io/gpuworks GitHub repository.2
3 Its pitch is a small control plane, mixed-accelerator management and model deployment without requiring Kubernetes.
1
3
What GPUWorks is designed to do
GPUWorks aims to bring accelerator resources into a shared cluster and provide a way to deploy and manage inference services. The project describes support for mixed accelerators across public and private clouds; a separate report names NVIDIA, Huawei Ascend, T-Head PPU and Hygon hardware.3
7 The available sources do not substantiate AMD support, so it should not be assumed from these claims.
The project emphasizes a small footprint: ZStack’s launch coverage describes a package larger than 40MB, one-command installation and no Kubernetes requirement. The GitHub project describes the download as about 40MB.1
3 Those figures are approximate descriptions, not a guarantee of the size of every installation or deployment.
The launch report also describes a catalog of nearly 400 models and more than 1,000 model-and-chip launch configurations, along with an embedded AI assistant for fault diagnosis and scheduling decisions that show their rationale.1 These are product claims in launch coverage; the supplied sources do not independently test the catalog or assistant.
GPU sharing and reported scale
GPUWorks describes using sleep and wake behavior so models can share GPU resources when they are not all needed at once. Its documentation says an instance can release GPU memory through the inference engine’s sleep mode while the container remains running, then be woken when needed.27 The announcement similarly presents model time-sharing as a way to reuse a pool of GPUs.
1
ZStack’s release reports a physical-hardware test in which one GPUWorks server managed 1,003 GPU servers, started more than 500 inference replicas in four minutes and delivered 69,000 tokens per second. The management server reportedly used nine CPU cores and 1.6GB of memory.2 These are vendor-reported test results, not independently verified benchmarks; the available material does not provide enough test detail to assess how they would translate to other hardware or workloads.
Open source and what the sources don’t establish
The Apache-2.0 announcement and public GitHub repository establish the project’s license and access point.2
3 The sources provided here do not document a specific contribution workflow or confirm every feature listed in broader descriptions of GPUWorks. In particular, they do not adequately establish AMD support, the full range of deployment environments, an OpenAI-compatible API, detailed billing and usage accounting, or the claimed model wake-up and comparative cold-start results.
The strongest supported takeaway is narrower: GPUWorks is an open-source, roughly 40MB platform intended to manage mixed-accelerator infrastructure and AI inference deployments without Kubernetes. Its headline scale figures and operational capabilities should be read as launch claims unless independently validated.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "GPUWorks: A Lightweight, Open-Source Platform for AI Inference"?
ZStack released GPUWorks under Apache 2.0 on November 4, 2026. The roughly 40MB platform is designed to manage mixed accelerator clusters without Kubernetes; its reported scale benchmarks come from the company, not in...
Sources
globenewswire.comA 40MB Platform to Manage a Thousand GPU Servers: GPUWorks Is Now Open Source
github.comGitHub - flitai-io/gpuworks: A ~45 MB MaaS platform: one cluster across public and private clouds, on any mix of accelerators.
geekpark.net40+ MB，纳管千台 GPU 服务器！GPUWorks 正式开源！ | 极客公园
pandaily.comGPUWorks Open-Sources 40MB Inference Platform Running Ascend, PPU, Hygon and NVIDIA in One Cluster