Eight gigabytes is the official minimum, not a generous shared-server target. Large imports, multiple users, PostgreSQL and other ZimaOS apps can leave little memory headroom, making 16 GB the better long-term class.
For larger annotation projects, multiple active users and all-in-one home servers.Label Studio Hardware Requirements: RAM, CPU, Storage & ML
Learn Label Studio hardware requirements for RAM, CPU, storage, PostgreSQL, Docker, datasets, ML backends and ZimaOS hardware.
Label Studio hardware requirements at a glance
Label Studio is a data-labeling web application, so its base hardware needs are driven by the application, database and dataset size rather than GPU inference. Current official guidance is substantially heavier than many lightweight cloud-template examples.
- CPU
- Current Label Studio documentation does not publish one universal minimum CPU core count. CPU demand grows with imports, exports, project size, concurrent annotators and any preprocessing or ML backend connected to the labeling server.
- RAM
- Current official server guidance says to use a minimum of 8 GB RAM, with 16 GB recommended. This official requirement takes precedence over third-party 1–2 GB cloud or Raspberry Pi examples.
- Storage
- Current official guidance recommends 50 GB disk for production. As a benchmark, 1 million labeling tasks occupy about 2.3 GB in SQLite, while source images, audio, video, exports and database growth can require substantially more.
- Database
- Current software requirements support PostgreSQL 13+ or SQLite 3.35+. Docker Compose uses PostgreSQL, which is the better long-term choice as projects, users and annotation volume grow.
- GPU
- Label Studio itself does not require a GPU. A GPU becomes relevant only when you connect a separate machine-learning backend that needs one; for example, Label Studio's SAM2 integration notes that SAM2 runs on GPU.
- Best Zima starting point
- ZimaBoard 2 832 reaches the official 8 GB RAM minimum but needs added SSD capacity for serious use. ZimaBoard 2 1664 reaches the official 16 GB recommendation; ZimaCube 2 Pro is the more balanced choice for larger datasets and multi-user labeling.
From official requirements to the right setup
Label Studio sizing starts with the official 8-GB/16-GB memory guidance, then adds dataset storage, database choice, annotator concurrency and optional ML backends.
-
Official requirements
Start with memory: use at least 8 GB RAM and target 16 GB for a more comfortable server. Do not size a ZimaOS deployment from third-party 1–2 GB examples that sit below current upstream guidance.
-
Confirm your needs
Plan persistent storage from the dataset rather than the application image. Label Studio recommends 50 GB for production and notes that 1 million SQLite labeling tasks use about 2.3 GB before counting source media and exports.
-
Leave room to grow
Choose SQLite only for smaller/simple deployments and move to PostgreSQL 13+ as project size and concurrency grow. The official Docker Compose path already uses PostgreSQL.
-
Run it on ZimaOS
Treat any ML backend as a separate compute service. If you connect SAM2, object detection, OCR or another model, add that model's CPU/GPU/RAM/VRAM requirements independently from Label Studio itself.
Check every playback client
- 8 GB minimum versus 16 GB recommended system memory
- Dataset size and source media type
- SQLite or PostgreSQL 13+ database
- Number of annotators and concurrent projects
- Image, audio, video or time-series workload
- Import/export and annotation-task volume
- Separate ML backend and GPU requirements if used
- Other ZimaOS containers, NAS and backup workloads
Official minimum requirements
Label Studio's current official documentation publishes explicit RAM and production-storage guidance, but not one universal CPU core-count minimum.
Use 8 GB RAM as the current official minimum, 16 GB as the current recommendation and 50 GB disk as the production storage recommendation. Third-party 1–2 GB deployments demonstrate that smaller experiments can start, but they should not replace the upstream server baseline.
| Requirement | Official minimum | What this supports |
|---|---|---|
| RAM | 8 GB minimum; 16 GB recommended | This is the current official Label Studio server guidance. |
| Production disk | 50 GB recommended | Disk usage scales with labeled data. Label Studio gives approximately 2.3 GB for 1 million labeling tasks in SQLite as a benchmark. |
| CPU | No universal official minimum published | Current server requirements do not specify a required processor model or core count. |
| Database | PostgreSQL 13+ or SQLite 3.35+ | The official Docker Compose deployment starts Label Studio with a PostgreSQL backend. |
| Network port | Port 8080 by default | Label Studio expects port 8080 to be open unless a different port is configured. |
| GPU | Not required by the base Label Studio application | GPU requirements belong to connected ML backends. Label Studio's official SAM2 example, for instance, requires GPU execution. |
When to upgrade your hardware
Upgrade Label Studio hardware when dataset scale, database activity or annotator concurrency becomes the bottleneck. A GPU upgrade only makes sense when an attached ML backend independently needs it.
Projects grow beyond the comfort of an 8 GB host
Dataset and annotation storage approach the system disk limit
A machine-learning backend is added for pre-annotation
The official 50-GB production recommendation only covers a starting capacity target. Source images, audio/video, exports and backups can exceed annotation-database size by a large margin.
For computer-vision, speech, video and large multi-project datasets.Label Studio can connect to a separate ML backend, but that model has its own compute requirements. SAM2, for example, requires GPU according to Label Studio's official integration tutorial.
For users adding active learning, auto-labeling, segmentation, OCR or model-assisted annotation.Plan hardware growth with confidence
Label Studio scales most cleanly when application/database state, source datasets and optional ML inference are treated as separate resource layers.
Move persistent data to SSD
The database, task metadata, uploads and exports generate active reads and writes. SSD storage improves responsiveness and gives more predictable behavior than relying indefinitely on small onboard eMMC.
Use SATA or NVMe SSD for Label Studio application/database data and active datasets.Use PostgreSQL as projects and users grow
SQLite is supported, but current official documentation also supports PostgreSQL 13+ and the Docker Compose path uses PostgreSQL by default.
Reserve SSD IOPS, RAM and backup capacity for PostgreSQL when the deployment becomes shared or long-lived.Separate source-media storage from application storage
Annotation metadata can be relatively small while source images, audio and video dominate capacity. Keeping them on separate storage makes expansion and backups easier.
Use ZimaCube 2 multi-drive storage when large media datasets become a central part of the workflow.Keep ML inference on a separately sized backend
Label Studio's ML backend runs as a separate service. This lets the labeling UI/database stay on stable server hardware while GPU-heavy models run elsewhere.
Use a remote GPU workstation or Creator Pack only when the actual ML model requires dedicated acceleration.Can it run on ZimaOS?
Label Studio is currently listed in the ZimaOS App Store under AI. ZimaOS can host the labeling application and persistent data, while optional ML backends can remain separate.
Install Label Studio from the ZimaOS App Store
ZimaOS currently lists Label Studio as an open-source data labeling application for audio, text, images, video and time-series data.
Open Label Studio in the ZimaOS App StoreUse the upstream 8 GB RAM minimum for sizing
The current Label Studio server documentation specifies at least 8 GB RAM and recommends 16 GB. ZimaOS packaging should not be used to lower that upstream application requirement.
Read Label Studio system requirementsAdd SSD storage before production-scale annotation
Label Studio recommends 50 GB for production and datasets grow independently from the application. Mount persistent Label Studio data on storage with substantial free capacity.
Read Label Studio installation guidanceChoose Zima hardware for your Label Studio workload
Label Studio's current 8 GB official minimum means ZimaBoard 2 832 and ZimaCube 2 Standard can run the application, but 16 GB configurations are the better long-term fit for shared projects, PostgreSQL and larger datasets.
Is this a personal/light Label Studio deployment or a larger shared annotation server?
Personal/light — smaller projects and limited concurrency
ZimaBoard 2 832 meets the current official 8 GB minimum, but add SSD storage because its 32 GB eMMC sits below the 50-GB production recommendation. Move to 1664 for the official recommended 16 GB memory level.
- Small personal labeling server with added SSDZimaBoard 2 832
- 16 GB recommended memory class for more headroomZimaBoard 2 1664
Shared/heavier — larger datasets, PostgreSQL or multiple users
ZimaCube 2 Pro is the more balanced stock option because it pairs 16 GB RAM with stronger CPU and 256 GB system storage. Creator Pack is only justified when a separate GPU ML backend is part of the same server.
- Storage-heavy but minimum-memory labeling platformZimaCube 2 Standard
- Balanced 16 GB shared Label Studio serverZimaCube 2 Pro
- Label Studio plus independently justified GPU ML backendZimaCube 2 Creator Pack
This is a workload guide, not a guaranteed annotator-count or task-throughput benchmark. Results depend on Label Studio version, dataset type, project size, database, concurrent users, storage latency, ML backends and other ZimaOS services.
| Zima hardware | Best for | Example workload | Core configuration | Recommended boundary | Next step |
|---|---|---|---|---|---|
| ZimaBoard 2 832 | Personal Label Studio projects that need a compact always-on server and can add external SSD storage. | Small image/text projects, light annotation, SQLite or modest PostgreSQL use and limited concurrent users. |
|
Eight GB only meets the official minimum, and 32 GB eMMC is below the official production disk recommendation. Do not present the stock 832 as a comfortable production setup without added storage. | Get Now |
| ZimaBoard 2 1664 | A compact Label Studio server that reaches the official 16 GB recommended memory level. | Larger personal projects, PostgreSQL, more concurrent annotation and several co-hosted ZimaOS services. |
|
64 GB eMMC only modestly exceeds the 50-GB production recommendation before source media and backups are counted. Use external SSD storage for serious projects. | Get Now |
| ZimaCube 2 Standard | Storage-heavy Label Studio projects where large media datasets matter more than additional system RAM. | Image, audio and video annotation with integrated multi-drive source-data storage and light-to-moderate concurrency. |
|
Standard only meets the 8-GB minimum rather than the 16-GB recommendation. ZimaBoard 2 1664 may be a better fit when memory, not storage, is the main constraint. | Get Now |
| ZimaCube 2 Pro | The most balanced current Zima configuration for larger Label Studio projects, PostgreSQL and shared annotation workloads. | Multiple projects, larger datasets, PostgreSQL, more concurrent annotators, exports and broader ZimaOS services. |
|
Do not interpret 16 GB as a guarantee for a specific annotator count. Very large projects or heavy ML backends may need more memory or separate compute. | Get Now |
| ZimaCube 2 Creator Pack | Label Studio combined with a GPU-dependent ML backend, large local datasets and other AI workloads on one server. | Annotation plus SAM2-style segmentation, model-assisted labeling or other compatible NVIDIA-accelerated ML services. |
|
Extreme overkill for Label Studio alone. Choose Creator Pack only when the connected ML model independently needs GPU acceleration, and verify that model's exact VRAM/software requirements. | Get Now |
What the Press Says
Highlights from trusted reviewers worldwide.
“ZimaCube 2: Not just another NAS, tested with 25TB storage, local AI agents, 4K transcoding, and real homelab workflows.”Read full review
“The ZimaBoard 2 is a compact x86 server board that can be turned into a mini NAS, home server, media box, or self-hosting hub.”Read full review
“ZimaCube 2: A modern, high-performance NAS with plenty of room to grow—built for users who want more than basic storage.”Read full review
“Coverage focused on ZimaCube 2's open hardware foundation, no monthly fee, and self-hosting flexibility.”Read full review
Loved by the Community
Stories and reviews from people who build with Zima every day.
★★★★★
Zima Blade Little yet Powerful
Maybe I am not digital natives but I live with PCs since 12 years old in 1984 when IBM PC clone come to my home. Many years have passed and many operating system I've tried. For me Zima blade and CasaOS was a quantum leap for home PC enthusiast and server lab machine to make me stay curious and relevant for this era.
★★★★★
Very good!!
I use ZimaCube Pro as 5th Proxmox cluster node. It runs several VMs and containers, including a VM with GPU passthrough to run a self-hosted LLM. A specific LXC container runs a Samba server for NAS capabilities using four of six RAID 6 SATA HDDs with ZFS.
★★★★★
Great innovation for mini server!
It is very useful and makes a powerful mini server for many purposes, including university and college students in engineering and electronics. Thank you so much for making this server.
★★★★★
Avaliação ZimaBoard 2
Construí um servidor de uso pessoal. O desempenho está muito bom e funciona perfeitamente onde quer que eu esteja. A surpresa é não dependermos de grandes estruturas para termos nosso próprio servidor de dados. Como iniciante, estou gostando bastante do ZimaOS, pois ele é simples e eficiente.
Frequently asked questions
These answers use current Label Studio server requirements as the baseline and explain why lightweight Railway or Raspberry Pi examples should not replace upstream guidance.
How much RAM does Label Studio need?
Current official Label Studio server guidance says to use at least 8 GB RAM and recommends 16 GB. This is the correct baseline for ZimaOS hardware planning.
Is 8 GB RAM enough for Label Studio?
Yes, it meets the current official minimum. It is best treated as a light/personal starting point rather than a generous production target. PostgreSQL, larger projects, multiple annotators and other ZimaOS services can make 16 GB the better choice.
Why does Railway say Label Studio only needs 1 GB RAM?
Railway's deployment page lists 1 vCPU/1 GB RAM/5 GB storage as its template minimum and 2–4 GB RAM as recommended, but this conflicts with current Label Studio upstream server guidance of 8 GB minimum and 16 GB recommended. The upstream requirement takes precedence.
How much storage does Label Studio need?
Current official guidance recommends 50 GB for production. Label Studio also gives a benchmark of about 2.3 GB for 1 million SQLite labeling tasks, but source media, exports, database files and backups can consume much more.
Can ZimaBoard 2 run Label Studio?
Yes. ZimaBoard 2 832 meets the official 8-GB RAM minimum, but its 32-GB eMMC is below the 50-GB production storage recommendation, so add SSD storage. The 1664 reaches the official 16-GB recommended memory level.
Does Label Studio need a GPU?
No for the base application. A GPU is only needed when a connected ML backend requires one. Label Studio's official SAM2 integration explicitly notes that SAM2 runs on GPU, which is a model-backend requirement rather than a Label Studio UI/database requirement.
Can Label Studio really run on a Raspberry Pi with 2 GB RAM?
A 2025 Peppe8o tutorial demonstrates Docker Label Studio on a Raspberry Pi 5 2GB and says 64-bit Raspberry Pi boards can run it. That is useful experimental evidence, but it sits below current upstream 8-GB server guidance and should not be used as the standard ZimaOS sizing baseline.
When should I choose ZimaCube 2 Pro instead of ZimaBoard 2 for Label Studio?
Choose Pro when you need the official 16-GB recommended memory level plus stronger CPU, 256-GB system storage, multi-drive datasets and more shared-server headroom. ZimaBoard 2 remains the better compact choice for smaller projects.
What sources and further reading informed this Label Studio hardware guide?
The two official Label Studio installation pages are the primary authority and currently specify 8 GB RAM minimum, 16 GB recommended, 50 GB production disk, PostgreSQL 13+ or SQLite 3.35+, and about 2.3 GB per 1 million SQLite labeling tasks. Railway's 1-GB/5-GB template is retained as a lightweight deployment example but conflicts with the upstream server requirement, so it is not used as the baseline. Labellerr explains installation methods and supported data types but does not provide numeric hardware minima. Peppe8o demonstrates a Raspberry Pi 5 2GB Docker deployment and at least 16 GB microSD, which is useful for showing experimental low-end viability but remains below the official server recommendation.
