OpenDataLab
opendatalab.com · Storage · 1 record · JSON
OpenDataLab (opendatalab.com / opendatalab.org.cn) Storage
caveatemailno cardphoneIper-object-cap
| Free tier | Not published. What is documented is capability rather than quota: the dataset center accepts file, folder, or archive uploads and auto-extracts archives server-side; public datasets require a license field before they can be published (so publishing is a two-step publish flow, not a raw PUT). OpenDataLab is the org behind the WanJuan corpus and mirrors COCO and similar multi-hundred-GB sets, so its scale class is genuinely multi-TB on the read side. unit TB; metered on unstated |
|---|---|
| Price | none stated |
| Per-object cap | No per-file cap, retention, or rate limit published |
| Retention | not stated |
| Egress | Free for public datasets; anonymous downloads. No egress line found. Served from Chinese CDN endpoints. |
| Identity | email email + likely +86 phone (Chinese platform). |
| Limits |
Unverified quota. Known: public publication requires a license to be set (a metadata gate, not a storage gate); archives are auto-extracted, so extracted size counts. No per-file cap, retention, or rate limit published. |
| Binding | per-object-cap the per-object cap sets the shard size and therefore the object count; it bounds the total more tightly than any headline quota |
| Interface | s3-compatible, rest, cli rest |
| Tooling | Same openxlab toolchain (OpenDataLab runs on the OpenXLab 浦源 docs stack): pip install openxlab, openxlab dataset upload-file/-folder. Also opendatalab/opendatalab-python-sdk (60 stars) and an OpenDataLab CLI documented at opendatalab.com/docs/developers/数据集/数据集CLI. There is no S3 API. |
| Composes with | Per-dataset repos, unlimited count, free account. 50 GB-class datasets are the platform's normal unit, so N datasets x 50-200 GB. |
| Reliability | API endpoints on opendatalab.com timed out from this network (25 s, then 30 s) while HTML pages served fine — treat the programmatic surface as CN-network-dependent. |
| Notes | Not published. What is documented is capability rather than quota: the dataset center accepts file, folder, or archive uploads and auto-extracts archives server-side; public datasets require a license field before they can be published (so publishing is a two-step publish flow, not a raw PUT). OpenDataLab is the org behind the WanJuan corpus and mirrors COCO and similar multi-hundred-GB sets, so its scale class is genuinely multi-TB on the read side. |
| Evidence rung | inferred One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-20 |
Evidence (3)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | opendatalab.com/docs/datasets/上传数据集.html | read live: upload types 文件/文件夹/压缩包, '如果您选择压缩包，系统将会在上传后自动解压您的文件', CLI and SDK examples |
| 2026-10-21 | opendatalab.com/docs/datasets/数据集创建流程.html | read live: publishing requires setting visibility to public and '公开数据集需具备license信息' |
| 2026-10-21 | opendatalab.com/api/v1/datasets?per=3 | probe: GET -> 404 (18 bytes); /api/v2/... -> connection timeout at 25 s |
Unknown or withheld values (4)
$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted