fshy / Compute / Netdisk and pan / Storage

Baidu

baidu.com · Compute, Netdisk and pan, Storage · 8 records · JSON

Baidu AI Studio 星河社区 (aistudio.baidu.com) — dataset + notebook storage Storagecaveatcountry-phoneno cardphoneRper-object-cap
Alibaba Cloud, Tencent Computefails: identitynational-idno cardRretention-clamped
Baidu AI Studio Computefails: identitynational-idno cardRvolume-quota
Baidu PaddlePaddle AI Computefails: identitynational-idno cardDrequest-rate
Baidu Pan (百度网盘) Netdisk and panfails: quotacountry-phoneno cardphoneRper-ip-shaping
Baidu Qianfan AppBuilder knowledge bases Storagefails: quotaany-phoneno cardphoneRper-object-cap
baidu qianfan 'unlimited' knowledge base storage Storagefails: quotaemail?card?R
baidu cloud ccr personal edition Storagedeademail?card?R

Baidu AI Studio 星河社区 (aistudio.baidu.com) — dataset + notebook storage Storage

caveatcountry-phoneno cardphoneRper-object-cap

aistudio.baidu.com · model-hub · storage/baidu-ai-studio-dataset-notebook-storage · JSON · compare

Free tier	Two separate pools, which is the interesting part: (1) Notebook persistent workspace — verbatim 'Notebook每个用户提供 100G 的云端持久化存储空间作为个人工作区，访问路径为 /home/aistudio/work' (100 GB per user, persisted in real time); (2) Datasets — '单个文件大小要小于50GB，一个数据集最多可上传50G数据内容' (single file under 50 GB, one dataset holds at most 50 GB of content). So 100 GB standing + 50 GB per dataset, and datasets are a separately-created object, not a folder in the workspace. unit GB; metered on unstated
Price	none stated
Per-object cap	50 GB per file (50000000000 bytes)
Retention	not stated
Egress	Free for public datasets — downloads are anonymous HTTP from Baidu's CDN with no egress line and no per-GB charge anywhere in the docs. No metered-egress surface exists because AI Studio is not a cloud-storage product with a billing meter.
Identity	country-phone email + any-phone, but the platform is China-facing and prior rounds recorded Chinese pans demanding +86 real-name verification and Alipay. Treat as country-phone until a registration attempt proves otherwise.
Limits	per-object-cap 50 GB per file source write-gated review purposes source 50 GB per file, 50 GB per dataset, 100 GB persistent workspace. Notebook main file (.ipynb) capped at 10 MB for preview purposes. Free GPU is 12 h/day with an 8-point/day sign-in economy, but compute is irrelevant here. Dataset uploads above 150 MB must go through the dataset path rather than direct project upload. No documented retention/expiry on datasets.
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, git, rclone, cli rest
Tooling	First-party pip package aistudio-sdk: `aistudio upload --repo-type {model,dataset} --include ... --max-workers N` with a User Access Token from aistudio.baidu.com/account/accessToken. Also git+LFS, and a web uploader. Python SDK + CLI both documented. No rclone backend, no S3 API.
Composes with	This is the multiplier case §1 describes: 50 GB is per DATASET, and the quota field explicitly separates the 100 GB workspace from dataset storage. N datasets x 50 GB is the intended pattern (people mount several datasets into one project simultaneously — Q21 '添加数据集' shows multiple datasets attached to a single project). 20 TiB = ~410 datasets. Whether the dataset COUNT is capped is the one number I could not establish — no page states a limit on number of datasets, and the platform's own docs describe creating datasets repeatedly as normal workflow.
Reliability	Datasets are mutable (versions exist — '生成新版本'). Notebook containers are recycled when idle and have a forced maximum lifetime, but /home/aistudio/work is described as real-time persistent, so the workspace is the durable part. Heavily China-network-oriented; download speeds from outside CN are untested here.
Notes	Two separate pools, which is the interesting part: (1) Notebook persistent workspace — verbatim 'Notebook每个用户提供 100G 的云端持久化存储空间作为个人工作区，访问路径为 /home/aistudio/work' (100 GB per user, persisted in real time); (2) Datasets — '单个文件大小要小于50GB，一个数据集最多可上传50G数据内容' (single file under 50 GB, one dataset holds at most 50 GB of content). So 100 GB standing + 50 GB per dataset, and datasets are a separately-created object, not a folder in the workspace.
Evidence rung	reported One page read moves it to documented.
Checked	2026-10-21 (inherited); stale on 2026-11-04
Parsed numbers	`$.limits[0].value` limit:per-object-cap 50.0 GB

Evidence (3)

date	source	what it proved
2026-10-21	ai.baidu.com/ai-doc/AISTUDIO/5k39vd65f	read live (via ?__provider=firecrawl after the default reader 504'd): Q7 '100G的云端持久化存储空间作为个人工作区，访问路径为 /home/aistudio/work/。写到该目录下的文件将被实时持久化存储'; Q1 '单个文件大小要小于50GB，一个数据集最多可上传50G数据内容'
2026-10-21	ai.baidu.com/ai-doc/AISTUDIO/nme7yirqi	read live: five upload methods; `aistudio upload -h` showing --repo-type {model,dataset} and --max-workers; token at /account/accessToken
2026-10-21	aistudio.baidu.com/datasetoverview	probe: GET -> 200 (13,685 bytes); GET https://aistudio.baidu.com/api/v1/datasets?per=2 -> 302 (API versioned/redirected, no anonymous JSON)

Unknown or withheld values (4)

$.free_tier.notes process
$.per_object_cap.notes not-extracted
$.retention.notes not-extracted
$.retention.stated not-extracted

Alibaba Cloud, Tencent Compute

fails: identitynational-idno cardRretention-clamped

cloud.baidu.com/campaign/free/index.html · cloud VM · compute/alibaba-cloud-tencent · JSON · compare

Free tier	Free trials of 7 to 30 days depending on vendor: Alibaba 7 days (14 on some products), Tencent 15 days, Huawei 30 days, Baidu 7 days with a 180-day campaign variant, JD 15 days. Baidu's rules page requires a real-name verified account: "users who register and complete real-name verification". unit none; metered on wall-clock; trial-credit
Price	none stated
Per-object cap	not applicable
Retention	Every one of the five gates the trial on real-name verification, which needs a national identity document, and every one ends within 30 days with no renewal path; the resource is auto-released at expiry.
Egress	not stated
Identity	national-id registration plus real-name verification before any trial: Baidu's campaign rules require "users who register and complete real-name verification" and its 180-day event requires verification within a week of registration, and JD Cloud requires new-user status plus real-name verification; the comparison read adds Tencent requiring a corporate check or a personal credit score of at least 600 and Huawei requiring a bank card bound to the account. The record's own limits row reads that verification as needing a national identity document, which is why the kind is national-id; no evidence row quotes the document list, so whether a foreign passport can substitute is not stated. No hold amount is stated and debit or prepaid acceptance is not stated
Limits	retention-clamped Every one of the five gates the trial on real-name verification, which needs a national identity document, and every one ends within 30 days with no renewal path; the resource is auto-released at expiry. source Every one of the five gates the trial on real-name verification, which needs a national identity document, and every one ends within 30 days with no renewal path; the resource is auto-released at expiry.
Binding	retention-clamped Every one of the five gates the trial on real-name verification, which needs a national identity document, and every one ends within 30 days with no renewal path; the resource is auto-released at expiry.
Interface	unknown inferred from the tooling the provider documents; not probed
Tooling	not stated
Composes with	Nothing that survives a month. The only durable Chinese free lane is the 180-day Baidu campaign, and it is annual and date-boxed rather than recurring.
Reliability	no reliability event recorded
Notes	This is one consolidated finding rather than five because the decision is the same and the door is the same: national identity. Two corrections worth carrying: JD's free period shrank from 6 months to 15 days, and Tencent's corporate requirement blocks an individual without a company. The 180-day Baidu figure is the largest in the group and still fails on identity. An account with a Chinese national ID would find a real 2 vCPU class trial here.
Evidence rung	reported
Checked	2026-10-22 (live); stale on 2026-11-05

Evidence (4)

date	source	what it proved
2026-10-22	cloud.baidu.com/campaign/free/index.html	campaign rules require registration plus real-name verification, cap each user to one free claim per product class, and state that resources are released at expiry, quoted from the rules block
2026-10-22	bce.baidu.com/event/clouday/index.html	a 180-day trial event restricted to accounts registered on one day and real-name verified within a week, after which resources are cleaned up 7 to 30 days post-expiry
2026-10-22	www.henghost.com/news/article/503871/	comparison table of the five vendors giving 7 to 30 day trial lengths and naming the identity conditions, including Tencent requiring a corporate check or a personal credit score of at least 600, and Huawei requiring a bank card bound
2026-10-22	github.com/vtukh186/jd-cloud-server-trial	JD Cloud trial is now 15 days rather than the 6 months older posts claim, and requires new-user status plus real-name verification

More fields (1)

{
 "wave2": {
  "class": "refutation-identity-gated",
  "free_tier_kind": "trial-credit",
  "shape": null
 }
}

Unknown or withheld values (8)

$.egress.metered egress-unknown
$.egress.terms not-stated
$.interface interface-unknown
$.per_object_cap.notes process
$.per_object_cap.stated not-applicable
$.reliability_note none-observed
$.retention.notes process
$.tooling not-stated

Baidu AI Studio Compute

fails: identitynational-idno cardRvolume-quota

aistudio.baidu.com/ · notebook runtime · compute/baidu-ai-studio · JSON · compare

Free tier	8 points a day, where 'each point buys 2 hours of the advanced GPU V100 16GB or 1 hour of the advanced GPU V100 32GB', so the daily allowance is 8 hours on a V100 32GB or 16 hours on a V100 16GB; the same documentation caps the advanced GPU at 48 hours a week at bronze level unit hours; metered on wall-clock; allowance
Price	none stated
Per-object cap	not applicable
Retention	the daily window is somewhere between 2.6 and 8 hours and the container is destroyed rather than paused, so a checkpoint outside the persistent mount is lost each day; the Chinese real-name requirement is the harder wall.
Egress	not stated
Identity	national-id not published
Limits	volume-quota 48 hours a week on the advanced GPU at bronze level, which is below the 8 hours a day the daily points would otherwise allow source retention-clamped a running project is terminated 10 minutes after it stops being used source the daily window is somewhere between 2.6 and 8 hours and the container is destroyed rather than paused, so a checkpoint outside the persistent mount is lost each day; the Chinese real-name requirement is the harder wall.
Binding	volume-quota the weekly cap of 48 hours at bronze level binds before the daily allowance: 8 hours a day is 56 hours a week, which the weekly cap refuses
Interface	unknown inferred from the tooling the provider documents; not probed
Tooling	not stated
Composes with	A V100 32 GB is the only free card found that can hold a 13B model unquantised, and as a chain element it is the training half and never the serving half.
Reliability	no reliability event recorded
Notes	A V100 at 32 GB a day for years is a real perpetual allowance and worth recording even though the door is shut, because the number is what makes the door worth naming. The daily figure is 2.6, 3 to 4 or 8 hours depending on the poster, which is the signature of a shifting allocation; the range is recorded rather than a point.
Evidence rung	reported
Checked	2026-10-22 (live); stale on 2026-10-24
Parsed numbers	`$.limits[0].value` limit:volume-quota 48.0 hours / week `$.limits[1].value` limit:retention-clamped 10.0 minutes

Evidence (2)

date	source	what it proved
2026-04-27	linux.do/t/topic/2068176	quoted, translated: Baidu AI Studio gives 2.6 hours of V100 32 GB a day, is extremely unstable, and writing large files crashes the container [re-read 2026-10-22: fetch exit 0, token overlap 0.0; the original capture stands]
2026-10-22	ai.baidu.com/ai-doc/AISTUDIO/nk39v9kec	quoted, translated from the vendor help page: 'each point buys 2 hours of the advanced GPU V100 16GB or 1 hour of the advanced GPU V100 32GB', and 'running a project each day grants 8 points', so the published daily allowance is 8 hours on a V100 32GB, which is the top of the stored 2.6-to-8-hour range and not the bottom; the same page states the environment 'keeps running for 10 minutes' after the browser closes and 'after 10 minutes it will be terminated', which supports the destroyed-not-paused claim. Supports the quota and the retention; the real-name gate is not on this page and stays reported

More fields (1)

{
 "wave2": {
  "class": "free-gpu-notebook",
  "free_tier_kind": "allowance",
  "shape": {
   "vcpu": null,
   "ram_gib": null,
   "gpu_class": "V100",
   "gpu_count": 1,
   "vram_gib": 32
  }
 }
}

Unknown or withheld values (9)

$.egress.metered egress-unknown
$.egress.terms not-stated
$.identity_cost.detail process
$.interface interface-unknown
$.per_object_cap.notes process
$.per_object_cap.stated not-applicable
$.reliability_note none-observed
$.retention.notes process
$.tooling not-stated

Baidu PaddlePaddle AI Compute

fails: identitynational-idno cardDrequest-rate

ai.baidu.com/ai-doc/AISTUDIO/nk39v9kec · GPU cloud · compute/baidu-paddlepaddle-ai · JSON · compare

Free tier	" 8 GPU V100 16GB 16 " - 8 compute-card points are granted per day, and " GPU V100 16GB 2 GPU V100 32GB 1 ". unit hours; metered on wall-clock; recurring-allowance
Price	none stated
Per-object cap	not applicable
Retention	not applicable
Egress	not stated
Identity	national-id not published
Limits	request-rate the mainland phone plus real-name check binds before any quota: without the 16 V100-hours a day are unreachable, and behind it the PaddlePaddle-only image restricts what runs at all source the mainland phone plus real-name check binds before any quota: without the 16 V100-hours a day are unreachable, and behind it the PaddlePaddle-only image restricts what runs at all
Binding	request-rate the mainland phone plus real-name check binds before any quota: without the 16 V100-hours a day are unreachable, and behind it the PaddlePaddle-only image restricts what runs at all
Interface	rest inferred from the tooling the provider documents; not probed
Tooling	not stated
Composes with	480 V100-hours a month is the largest recurring GPU allowance found anywhere and it exceeds Kaggle's by a factor of nearly four it composes with nothing here because the door is a mainland real-name account.
Reliability	no reliability event recorded
Notes	This is the biggest number in the survey and it is unreachable for a user: registration is a mainland phone number and the account behind it carries a real-name check, and the platform's own third-party guides state the frameworks are restricted to the PaddlePaddle ecosystem. 480 hours a month is 16 x 30, arithmetic on the published daily grant. Two open items: the official page does not state how long the daily 8 points stay valid (one forum post says 24 hours, unverified), and the weekly ceiling in the levelling table (48 hours a week at bronze) may bind before the daily grant does at low account levels.
Evidence rung	documented
Checked	2026-10-22 (live); stale on 2026-11-21

Evidence (4)

date	source	what it proved
2026-10-22	ai.baidu.com/ai-doc/AISTUDIO/nk39v9kec	quoted: 8 GPU V100 16GB 16
2026-10-22	ai.baidu.com/ai-doc/AISTUDIO/nk39v9kec	quoted: GPU V100 16GB 2 GPU V100 32GB 1 ; billing per minute, aggregated every 6 minutes; the environment is kept alive 10 minutes after the browser closes
2026-10-22	ai.baidu.com/ai-doc/AISTUDIO/1kczvydw9	quoted progression table: Notebook GPU weekly hours rise with account level from 48 hours (bronze) to 120 hours (diamond), and GPU weekly hours from 4 to 24
2026-10-22	ai.baidu.com/ai-doc/AISTUDIO/nk39v9kec	probe 200, 836156 bytes, 2026-10-22T14:24:18Z, no false-200 signature

More fields (1)

{
 "wave2": {
  "class": "recurring-daily-quota",
  "free_tier_kind": "recurring-allowance",
  "shape": {
   "vcpu": null,
   "ram_gib": null,
   "gpu_class": "V100 16GB",
   "gpu_count": 1
  }
 }
}

Unknown or withheld values (9)

$.egress.metered egress-unknown
$.egress.terms not-stated
$.identity_cost.detail process
$.per_object_cap.notes process
$.per_object_cap.stated not-applicable
$.reliability_note none-observed
$.retention.notes process
$.retention.stated not-applicable
$.tooling not-stated

Baidu Pan (百度网盘) Netdisk and pan

fails: quotacountry-phoneno cardphoneRper-ip-shaping

pan.baidu.com · consumer-pan · netdisk-pan/baidu-pan · JSON · compare

Free tier	300 GB after a mobile-client login; 100 GB on registration alone: '用户注册即有100G云空间，在手机客户端登录则升级为300G' unit GiB; metered on stored-bytes
Price	none stated
Per-object cap	1 GB per file on the web uploader, 4 GB on the desktop client, 10 GB on VIP and 20 GB on SVIP (1000000000 bytes)
Retention	Trash holds deleted files for 10 days. The free allowance was cut from 2 TB to 100 GB in 2020 for accounts not logged in for two years
Egress	Metered by throttle rather than by byte: non-member downloads are capped around 0.1 MB/s and anything sizeable needs the desktop or mobile client
Identity	country-phone a Chinese mobile number. The open platform adds a developer app on top
Limits	per-ip-shaping about 0.1 MB/s on a non-member download source per-object-cap 1 GB per file on web, 4 GB on the desktop client, free source retention-clamped the free allowance drops to 100 GB for accounts idle for two years source
Binding	per-ip-shaping 300 GB is 60% of the filter and 0.1 MB/s is 8.6 GB a day, so the throttle binds first: the corpus would need 116 days of continuous reading to serve one day's target.
Interface	rest
Tooling	an official open platform with OAuth 2.0 at pan.baidu.com/union/doc, for example GET /api/quota with an access token returning total, free and used in bytes; the OpenList BaiduNetdisk driver; a Baidu developer app is required
Composes with	300 GB per Baidu account and Baidu accounts need a phone number. At 0.1 MB/s the non-member throttle makes even 500 GB a day impossible, so the read path fails independently of the quota
Reliability	Content scanning applies even to private files: matching files become unreadable for the uploader too. The free tier is effectively read-only at serving scale. The guest share path gets further than expected and then stops: /share/verify returns errno 0 with a randsk cookie and the page renders the real file tree, but /share/list answers errno 2 and every byte path needs BDUSS
Notes	One record for the provider and its two seed sub-findings, the 2 TB to 100 GB cut and the anonymous share path. The anonymous path is the more interesting of the two: it proves the share metadata is public and the bytes are not, which is the pattern the whole topic is built on.
Evidence rung	reported
Checked	2026-10-21 (inherited); stale on 2026-11-04
Parsed numbers	`$.limits[0].value` limit:per-ip-shaping 0.1 MB `$.limits[1].value` limit:per-object-cap 1.0 GB `$.limits[2].value` limit:retention-clamped 100.0 GB

Evidence (6)

date	source	what it proved
2026-10-21	zh.wikipedia.org/wiki/中国大陆网盘列表	table row '百度网盘, 300G, 无, 有, 有, 0.1M/s, 是, 是' and '注册即有100G云空间，在手机客户端登录则升级为300G'
2026-10-21	pan.baidu.com/disk/help	'普通用户使用百度网盘Web端上传文件时，单文件最大支持1G' desktop client 4 GB; VIP 10 GB; SVIP 20 GB
2026-10-21	www.ithome.com/0/511/746.htm	'若过去两年未登录过，免费空间将从 2TB 调整为 100GB'
2026-10-21	probe:POST https://pan.baidu.com/share/verify?surl=RK7uBqaqgqJHLJbadXI48g	200 with {"errno":0,"randsk":"oMkK%2Fo1NQiRi7U58..."}, so the passcode gate opens a guest session
2026-10-21	probe:GET https://pan.baidu.com/share/list?uk=858219170&shareid=47786905570	200 with {"errno":2,"show_msg":"啊哦，链接出错了"} for the same uk and shareid the page rendered
2026-10-21	probe:GET https://pan.baidu.com/rest/2.0/xpan/multimedia	{"errmsg":"Invalid Bduss","errno":-6}, so every byte path needs a logged-in session

More fields (5)

{
 "access_mode": "registration-gated",
 "anonymous_read": "metadata yes, bytes no",
 "aliases": [
  "Baidu Pan / 百度网盘"
 ],
 "cross_topic_record": "dossier/storage/baidu-pan.json",
 "cross_topic_note": "A record for this provider already exists under the storage topic; this one is scoped to its role as a pan, following the same split dossier/netdisk-pan/lanzou.json uses."
}

Baidu Qianfan AppBuilder knowledge bases Storage

fails: quotaany-phoneno cardphoneRper-object-cap

ai.baidu.com/ai-doc/AppBuilder/lmde5yr7s · Chinese agent platform / RAG knowledge base · storage/baidu-qianfan-appbuilder-knowledge-bases · JSON · compare

Free tier	Advertised as '文件存储和处理不限量' (file storage and processing unlimited) on the shared-storage path; documented default is 100 knowledge bases x 800 files each = 80,000 files. PDF up to 500 MB / 3,000 pages per file on one endpoint; txt 10 MB, docx/doc/ppt/pptx 50 MB on the ingestion path. unit objects; metered on stored-bytes
Price	none stated
Per-object cap	per-object-cap 500 MB (500000000 bytes)
Retention	not stated
Egress	No egress fee on platform-hosted knowledge files.
Identity	any-phone any-phone
Limits	per-object-cap per-object-cap 500 MB source per-object-cap 500 MB (PDF), 50 MB (other docs), 10 MB (txt). BOS-backed import is the documented route to '大规模数据' (large-scale data).
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, rest REST (Qianfan API: upload-to-knowledge-base, create dataset); the BOS-backed path is the only one with a real quota story
Tooling	Baidu Qianfan SDK / plain REST; BOS S3-like API for the source-import path
Composes with	100 KBs x 800 files = 80,000 files per account on shared storage; 'unlimited' only on the BOS/BES enterprise path, which means paying for BOS — that collapses to BOS pricing, which is out of scope.
Reliability	The 'unlimited' claim is 2024 marketing for the enterprise RAG tier and does not survive contact with the docs: the same help centre says the shared-storage path 'supports 小规模文件导入' (small-scale import), and the unlimited path runs through user-owned BOS which is billed. Baidu Cloud registration requires a phone number. Treat as a documented-name entry, not a qualifying provider.
Notes	Advertised as '文件存储和处理不限量' (file storage and processing unlimited) on the shared-storage path; documented default is 100 knowledge bases x 800 files each = 80,000 files. PDF up to 500 MB / 3,000 pages per file on one endpoint; txt 10 MB, docx/doc/ppt/pptx 50 MB on the ingestion path.
Evidence rung	reported One page read moves it to documented.
Checked	2026-10-21 (inherited); stale on 2026-11-04
Parsed numbers	`$.limits[0].value` limit:per-object-cap 500.0 MB

Evidence (5)

date	source	what it proved
2026-10-21	qianfan.cloud.baidu.com/qianfandev/topic/361340	'✨文件存储和处理不限量' (file storage and processing unlimited), 2024-07-22
2026-10-21	ai.baidu.com/ai-doc/AppBuilder/lmde5yr7s	'本地上传 \| 上传本地数据并导入 AppBuilder 共享存储服务，支持小规模文件导入' vs '百度对象存储（BOS）\| 导入 BOS 中的数据，支持导入大规模数据'
2026-10-21	cloud.baidu.com/doc/qianfan/s/umh4stn0t	'每位用户可创建100个知识库，每个知识库支持上传800个文件'
2026-10-21	ai.baidu.com/ai-doc/AppBuilder/Olz18mbpq	'.pdf文件不能超过500MB且不能超过3000页；其他类型文件不能超过50MB'
2026-10-21	www.yun88.com/qa/4409.html	'默认800个，打通BOS+BES后不设上限' (default 800, no cap once BOS+BES is connected)

More fields (1)

{
 "dead_end": {
  "failure_class": "quota",
  "why": "Advertised as '文件存储和处理不限量' (file storage and processing unlimited) on the shared-storage path; documented default is 100 knowledge bases x 800 files each = 80,000 files. PDF up to 500 MB / 3,000 pages per file on one endpoint; txt 10 MB, docx/doc/ppt/pptx 50 MB on the ingestion path.",
  "retry_when": null,
  "first_recorded": "2026-10-21"
 }
}

Unknown or withheld values (5)

$.extras.dead_end.retry_when process
$.free_tier.notes process
$.per_object_cap.notes not-extracted
$.retention.notes not-extracted
$.retention.stated not-extracted

baidu qianfan 'unlimited' knowledge base storage Storage

fails: quotaemail?card?R

qianfan.cloud.baidu.com/qianfandev/topic/361340 · rejected-provider · storage/baidu-qianfan-unlimited-knowledge-base-storage · JSON · compare

Free tier	The unlimited claim is enterprise-tier marketing from 2024 and routes through user-funded BOS. The shared-storage path is documented as '小规模文件导入' (small-scale import) with 800 files per KB a unit none; metered on none
Price	none stated
Per-object cap	not recorded
Retention	not recorded
Egress	not recorded
Identity	email (placeholder) not published
Binding	none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit
Interface	unknown
Tooling	not recorded
Composes with	not recorded
Reliability	The unlimited claim is enterprise-tier marketing from 2024 and routes through user-funded BOS. The shared-storage path is documented as '小规模文件导入' (small-scale import) with 800 files per KB a
Notes	The unlimited claim is enterprise-tier marketing from 2024 and routes through user-funded BOS. The shared-storage path is documented as '小规模文件导入' (small-scale import) with 800 files per KB a
Evidence rung	reported
Checked	2026-10-21 (inherited); stale on 2026-11-04

Evidence (3)

date	source	what it proved
2026-10-21	qianfan.cloud.baidu.com/qianfandev/topic/361340	The unlimited claim is enterprise-tier marketing from 2024 and routes through user-funded BOS. The shared-storage path is documented as '小规模文件导入' (small-scale import) with 800 files per KB a
2026-10-21	ai.baidu.com/ai-doc/AppBuilder/lmde5yr7s	cited as the basis of the rejection: The unlimited claim is enterprise-tier marketing from 2024 and routes through user-funded BOS. The shared-storage path is documented as '小规模文件导入' (small-scale import) with 800 files per KB a
2026-10-21	cloud.baidu.com/doc/qianfan/s/umh4stn0t	cited as the basis of the rejection: The unlimited claim is enterprise-tier marketing from 2024 and routes through user-funded BOS. The shared-storage path is documented as '小规模文件导入' (small-scale import) with 800 files per KB a

More fields (1)

{
 "dead_end": {
  "failure_class": "quota",
  "why": "The unlimited claim is enterprise-tier marketing from 2024 and routes through user-funded BOS. The shared-storage path is documented as '小规模文件导入' (small-scale import) with 800 files per KB a",
  "retry_when": null,
  "first_recorded": "2026-10-21"
 }
}

Unknown or withheld values (15)

$.chaining not-recorded
$.confidence_note withheld
$.egress.terms not-recorded
$.extras.dead_end.retry_when process
$.free_tier.notes process
$.identity_cost.detail defaulted
$.identity_cost.kind identity-defaulted
$.interface interface-unknown
$.interface_detail withheld
$.limits_note process
$.per_object_cap.notes not-recorded
$.per_object_cap.stated not-recorded
$.retention.notes process
$.retention.stated not-recorded
$.tooling not-recorded

baidu cloud ccr personal edition Storage

deademail?card?R

cloud.baidu.com/doc/CCR/s/jkw60wx3m · rejected-provider · storage/baidu-cloud-ccr-personal-edition · JSON · compare

Free tier	Dead since 2024-04-25: personal-edition instances no longer accept new namespaces or repositories at all, and no external image migration is allowed. Only pre-existing images keep working. unit none; metered on none
Price	none stated
Per-object cap	not recorded
Retention	not recorded
Egress	not recorded
Identity	email (placeholder) not published
Binding	none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit
Interface	unknown
Tooling	not recorded
Composes with	not recorded
Reliability	Dead since 2024-04-25: personal-edition instances no longer accept new namespaces or repositories at all, and no external image migration is allowed. Only pre-existing images keep working.
Notes	Dead since 2024-04-25: personal-edition instances no longer accept new namespaces or repositories at all, and no external image migration is allowed. Only pre-existing images keep working.
Evidence rung	reported
Checked	2026-10-21 (inherited); stale on 2026-11-04

Evidence (1)

date	source	what it proved
2026-10-21	cloud.baidu.com/doc/CCR/s/jkw60wx3m	Dead since 2024-04-25: personal-edition instances no longer accept new namespaces or repositories at all, and no external image migration is allowed. Only pre-existing images keep working.

More fields (1)

{
 "dead_end": {
  "failure_class": "dead-service",
  "why": "Dead since 2024-04-25: personal-edition instances no longer accept new namespaces or repositories at all, and no external image migration is allowed. Only pre-existing images keep working.",
  "retry_when": null,
  "first_recorded": "2026-10-21"
 }
}

Unknown or withheld values (15)

$.chaining not-recorded
$.confidence_note withheld
$.egress.terms not-recorded
$.extras.dead_end.retry_when process
$.free_tier.notes process
$.identity_cost.detail defaulted
$.identity_cost.kind identity-defaulted
$.interface interface-unknown
$.interface_detail withheld
$.limits_note process
$.per_object_cap.notes not-recorded
$.per_object_cap.stated not-recorded
$.retention.notes process
$.retention.stated not-recorded
$.tooling not-recorded