Alibaba Cloud PAI iTAG

Overview

Alibaba Cloud PAI iTAG is a managed platform for labeling image, text, video, audio, and multimodal data. It is best suited to teams already working with Alibaba Cloud OSS and needing configurable annotation and review workflows. Its range of task types is useful, but its current access restrictions make it a poor choice for new users.

Key features

iTAG offers built-in image tasks for classification, object detection, OCR, table recognition, and semantic segmentation. Official multimodal templates include image-text explanation, conversation ranking, visual question answering, and video OCR. Teams can also select tenant templates or create custom ones for business-specific labeling schemes. That breadth helps teams match annotation to different data and tasks, though custom work requires designing a suitable template.

A job combines an OSS-backed dataset with a template, then configures labels, subtasks, checks, labelers, and reviewers. The ability to include checks and reviewers makes the workflow more suitable for teams that need a review step than a simple one-pass labeling process. Selected large-model templates, including image-to-text and image-text explanation, are currently free; Alibaba Cloud says it will notify users if pricing changes.

The OpenAPI supports listing, creating, updating, and deleting templates and labeling tasks, which can help teams manage those resources programmatically. For API access, Alibaba Cloud recommends an API-only AccessKey for a RAM user with least-privilege permissions. The overview also points users to an FAQ and an Alibaba Cloud AI assistant for product, solution, and documentation questions.

Storage is a significant constraint: input data must be in Alibaba Cloud OSS, so local or NAS data needs to be uploaded first. The iTAG workspace and OSS bucket must be in the same region, and importing a dataset requires an index file rather than a direct link to a folder of raw files. Teams with data already organized in OSS are better placed to use iTAG; teams relying on local or NAS storage face extra preparation.

From August 1, 2026, iTAG 1.0 is allowlist-only and in maintenance mode. Existing customers with prior labeling tasks or datasets can continue using current features, but the overview directs new users seeking AI-powered automatic dataset labeling to Multimodal Data Management instead.

Pricing

iTAG manual labeling: free. The platform is free when a team labels manually, but OSS storage and data transfer are billed separately. This plan suits teams that can supply their own annotators and need to avoid a platform fee; the separate OSS costs still matter, and the plan is not a reason to overlook the OSS requirement. There is no free trial.

Selected large-model labeling templates are currently free as well, including image-to-text and image-text explanation. Alibaba Cloud says it will notify users if that pricing changes.

Platforms

iTAG is a managed cloud deployment with web, API, Linux, and Windows support. Its supported operating systems include Windows, Linux, Anolis OS, Aliyun Linux 2 and 3, Kylin, and Unity Operatin. Container support, an admin console, clustering, and the Jakarta EE standard are also listed.

Who it's for

iTAG is a better fit for existing Alibaba Cloud customers with OSS-hosted datasets, varied labeling needs, and a reason to configure human review. It is not a strong starting point for new users: access is restricted to the allowlist, and the product is in maintenance mode. Teams seeking AI-powered automatic dataset labeling should consider Multimodal Data Management instead.

Pros and cons

• Pro: Image, text, video, audio, and multimodal support gives teams room to handle varied data types in one labeling platform.
• Pro: Configurable labels, subtasks, checks, labelers, and reviewers support more structured workflows than a single annotator pass.
• Pro: Manual labeling costs nothing; selected large-model templates are also currently free.
• Con: Input data must be uploaded to OSS, and the bucket must be in the workspace's region.
• Con: Dataset import requires an index file, adding a preparation step for teams expecting to point directly to a raw-file folder.
• Con: Allowlist-only access and maintenance mode limit its suitability for new users.
• Con: OSS storage and data transfer are billed separately, so free platform use does not eliminate infrastructure costs.

Alternatives

AI Data Labeling Tools is a useful place to compare other labeling products. Choose doccano if a free, MIT-licensed option with web, API, and self-hosted platforms better fits your deployment needs. Label Studio is another freemium alternative with a free plan and trial, plus web, API, and self-hosted support. Choose BRAT if a free, MIT-licensed, self-hosted option is the better fit. Roboflow offers a free tier with 10 credits a month, stated as enough to train about 30 models or run 80,000 inferences; its Core plan is $39 a month. Amazon SageMaker Autopilot has a free trial and pay-as-you-go pricing, with no minimum fees or upfront commitments for on-demand use. Choose Argilla if free, open-source software deployable on Hugging Face Spaces or your own infrastructure is preferable. CVAT offers free community and online options, including a free online plan limited to 1 member, 1 project, 3 tasks, and 1 GB. Intellisane Annotation Platform has a free plan capped at 3 members, 2 projects, 1,000 annotations, and 200 assets.

Verdict

For existing Alibaba Cloud customers with OSS-hosted data who need varied annotation tasks and review workflows, iTAG’s breadth and free manual-labeling platform make it worth considering. The OSS setup adds friction, and allowlist-only access in maintenance mode is the decisive reason most new users should look elsewhere.