Alibaba Cloud PAI iTAG

Overview

Alibaba Cloud PAI iTAG is a managed platform for labeling image, text, video, audio, and multimodal datasets. It is best suited to teams whose data and workflows already sit in Alibaba Cloud, especially those with varied annotation needs and human review. Its range of templates is useful, but the OSS dependency and allowlist-only maintenance status make it a difficult choice for a new labeling program.

Key features

Image workflows cover 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 start from official or tenant templates, or create a custom template for a business-specific labeling scheme. That breadth can fit specialized projects, though custom templates require teams to define the structure that suits their work.

A job combines an OSS-backed dataset with a template, then configures labels, subtasks, checks, labelers, and reviewers. The workflow can support review and quality checks beyond a single annotation pass. Selected large-model templates, including image-to-text and image-text explanation, are currently free; Alibaba Cloud says it will notify users if their pricing changes.

The OpenAPI supports listing, creating, updating, and deleting templates and labeling tasks, making it relevant to teams that want to manage those resources programmatically. For API access, Alibaba Cloud recommends a RAM user with an API-only AccessKey and least-privilege permissions.

Storage is a central constraint: input data must be in Alibaba Cloud OSS, and local or NAS data must be uploaded there first. The iTAG workspace and bucket must share a region, and importing a dataset requires an index file rather than a folder path. This is a workable fit for teams already organized around OSS, but extra preparation for other storage setups.

Pricing

iTAG manual labeling: free. The platform is free when a team performs labeling manually, but OSS storage and data transfer are billed separately. This plan suits teams that want the labeling platform without a platform charge and can account for those separate cloud costs. It does not make the full workflow cost-free when data storage or transfer is involved.

Selected large-model labeling templates are also currently free. Alibaba Cloud says it will notify users if pricing changes. There is no free trial; the free manual-labeling plan is the entry point. From August 1, 2026, iTAG 1.0 is allowlist-only and in maintenance mode, so continued use is limited to existing customers with prior labeling tasks or datasets.

Platforms

iTAG is a managed cloud service with web and API access, and the platform lists Linux and Windows support. Its cloud deployment and OSS-backed datasets make it most natural for teams already using Alibaba Cloud rather than those seeking a self-hosted labeling system.

Who it's for

iTAG makes the most sense for existing Alibaba Cloud customers who need annotation across several data types, configurable templates, and reviewer participation. It is a poor fit for new users seeking an actively maintained starting point: iTAG 1.0 is allowlist-only and in maintenance mode, and Alibaba Cloud directs users seeking AI-powered automatic dataset labeling to its Multimodal Data Management feature.

Pros and cons

• Pro: Broad modality and task coverage, including image-specific work and official multimodal templates, can support varied labeling projects in one platform.
• Pro: Configurable jobs with checks, labelers, and reviewers accommodate quality control, while selected model-assisted templates are currently free.
• Con: Mandatory OSS storage, same-region requirements, and index-file imports add setup work for data held elsewhere.
• Con: Manual labeling is free at the platform level, but OSS storage and transfer are separate charges.
• Con: Allowlist-only maintenance status makes iTAG a poor foundation for new deployments.

Alternatives

AI Data Labeling Tools is a useful category page for comparing labeling options. Choose doccano instead if free, MIT-licensed software and broad platform availability matter more than iTAG's Alibaba Cloud workflow. Label Studio is another freemium option, with a free plan and free trial.

For self-hosted options, BRAT offers free MIT-licensed installation, while Argilla is free, open-source software that can run on Hugging Face Spaces or a team's own infrastructure. CVAT has a free MIT-licensed community plan for personal use and small teams, as well as an online free plan limited to one member, one project, three tasks, and 1 GB of storage.

Roboflow may suit teams that want a freemium option with a Free Tier that includes 10 monthly credits, described as enough for about 30 model trainings or 80,000 inferences. Intellisane Annotation Platform offers a free plan capped at three members, two projects, 1,000 annotations, and 200 assets. Amazon SageMaker Autopilot is a paid, pay-as-you-go alternative with a free trial; its on-demand pricing has no minimum fees or upfront commitments.

Verdict

Choose Alibaba Cloud PAI iTAG if you are an existing Alibaba Cloud customer who needs varied labeling templates, configurable review workflows, and can keep datasets in OSS. The strongest reason to pick it is the breadth of annotation tasks and human-review controls under a free manual-labeling platform charge. Look elsewhere if you are a new user or need a platform to build a new long-term workflow around: allowlist-only maintenance and the OSS requirements are substantial constraints.