Big Data & AI Platform Monthly Newsletter—May 2026

The Big Data & AI Product Monthly Newsletter for May 2026 covers technology updates, product and feature releases, market updates, and customer application practices to help you quickly understand the latest developments in Alibaba Cloud's Big Data & AI offerings.

I. Product Feature Releases

Platform for AI (PAI) — Deep Integration with LLM on Model Studio

By adding LLM on Modul Studio and deeply integrating them into PAI product modules, we provide users with more out-of-the-box model choices to meet the growing demand for model services.

MaxCompute — MaxCompute SQL Supports Scenario-Based AI Functions

MaxCompute SQL adds 6 new scenario-based AI Functions including data classification, extraction, sentiment analysis, and translation.

MaxCompute — MaxCompute SQL Supports UNNEST Operator

MaxCompute SQL now supports array expansion via the UNNEST operator, simplifying nested data processing. It also supports coordination with JOIN syntax via ON conditions for associative filtering, suitable for array split statistics, struct field extraction, and associative match filtering scenarios.

MaxCompute — MaxCompute SQL Supports Creating Temporary Tables in Script Mode

MaxCompute SQL script mode now supports temporary tables. Within the script execution lifecycle, you can create temporary tables via CREATE TEMPORARY TABLE to cache intermediate results for multiple reuses within the same script, reducing repeated query overhead.

MaxCompute — MaxQA Supports Automatic Elastic Upper Limit CU (Autoscale)

MaxQA supports automatic elastic upper limit CU (Autoscale). Interactive Quota types can be set with a step size of 25 CU, and the configurable automatic elastic upper limit CU value range is [0, Level-1 Quota AutoscaleLimitCU].

MaxCompute — MaxCompute Supports Managing Iceberg Tables

MaxCompute supports storing Apache Iceberg tables on Alibaba Cloud Object Storage Service (OSS), and manages metadata, permissions, and data lifecycle through MaxCompute for efficient querying, writing, and management. Iceberg tables are compatible with open-source engines like Spark, supporting multi-engine sharing of the same data, suitable for Lakehouse architecture.

MaxCompute — Blob Data Type and Multimodal Storage

With the Blob type, raw files, metadata, and annotation information of multimodal data can be uniformly stored in the same MaxCompute table, queried and maintained via SQL, and processed in batch through MaxFrame and SQL UDF.

MaxCompute — MaxFrame AI Function Adds Multimodal Data Processing Capability

MaxFrame AI Function further integrates with Alibaba Cloud Bailian Platform. Without the need to self-encapsulate UDFs or maintain DashScope Keys, you can directly call Model Studio's multimodal models within MaxFrame DataFrame expressions, supporting declarative invocation, automatic batch distributed inference, row-level fault tolerance, and precise reruns.

MaxCompute — MaxFrame Multimodal Operator Module Update

The MaxFrame multimodal operator module officially extends its capability boundary to audio, adding the Series.mf.audio accessor, covering the full pipeline of speech processing.

MaxCompute — MaxFrame Releases Intelligent Driving Industry Scenario Skills

For industry scenarios such as intelligent driving, smart cabin, and visual perception, MaxFrame officially releases the MaxFrame Intelligent Driving Video Processing Skill. It covers core pipelines including video frame extraction, keyframe labeling, image labeling/vectorization, and image table embedding appending.

MaxCompute — MaxFrame Apply / UDF Major Enhancement

Output dtypes can now be inferred via function type hints, eliminating the need to manually write dtypes; apply / apply_chunk can now capture raw failure data when errors occur, facilitating issue troubleshooting.

MaxCompute — MaxFrame Network Security Enhancement

Added with_network_options, supporting VPC network link configuration for private network environment access.

MaxCompute — MaxFrame DataFrame Lineage Capability Released

New MaxFrame DataFrame ↔ MaxCompute table column-level lineage tracking: covering aggregation / join / projection / selection / setitem / source / sink scenarios, viewable in DataWorks Data Lineage for data governance and impact analysis.

Hologres — Hologres Launches AI Assistant

Hologres introduces an AI Assistant powered by large models and intelligent Agent systems, providing comprehensive capabilities including Q&A support for Hologres, automated operations, performance optimization guidance, development assistance, and data analysis.

DataWorks — OpenAPI Free Quota Increased for Standard and Professional Editions, Pay-As-You-Go Beyond Limits

Provides higher OpenAPI free call quotas for Standard and Professional Edition customers, avoiding "stop on over-limit" and ensuring continuous API access with elastic scalability.

DataWorks — User-Exclusive AI Assistant Built on OpenClaw Framework Now Live

The AI Assistant service integrates with IM platforms such as DingTalk and Feishu, enabling natural language interactions for task diagnosis, alert analysis, and operations directly from the IM client — transforming from passive alert response to proactive intelligent operations.

DataWorks — Data Studio Supports AI-Powered Code Review Suggestion Summarization

During the code review process in data development, reviewers can leverage AI capabilities to generate summarized review suggestions for review tickets, significantly improving review efficiency.

DataWorks — Data Agent Upgrade with Three New Panels

DataWorks Data Agent adds three new panels: context timing waterfall chart visualizes the call chain, deliverables details supports right-side preview for code, documents, and images, and session environment centrally displays workspace information, comprehensively upgrading interaction experience and readability.

DataWorks — Notebook/PyODPS Nodes Support AI-Powered Lineage Analysis, Registration, and Management

For Python/MaxFrame code logic that involves MaxCompute tables, datasets, or OSS paths as inputs or outputs, AI can intelligently parse the data lineage and register it to the lineage chain with one click, effectively improving development efficiency.

DataWorks — Data Quality Adds Support for Multiple Quality Rule Templates Including JSON Format Validity, Field Consistency, Accuracy, and Timeliness

DataWorks Data Quality now adds multiple quality rule templates covering validity, consistency, accuracy, and timeliness. Users can perform quality validation for scenarios including JSON field format validation, intra-table field value comparisons, aggregated value fluctuation comparisons by specified dimensions and metric fields, and time-series data delay detection — enabling users to quickly implement metric-oriented data quality governance.

DataWorks — Data Integration Adds Support for Multiple Data Sources and Data Synchronization Capabilities

Addresses user requirements for expanded data source support and enhanced data synchronization capabilities within Data Integration tasks, enabling seamless data movement across more heterogeneous data sources.

Elasticsearch — Search Compute-Storage Separation DFS Supports Version 8.17 and Single AZ to Multi AZ Scaling

The search scenario compute-storage separation feature DFS now supports version 8.17, and single availability zone instances can be scaled to multi availability zones, improving search performance and high availability.

Milvus — External Collection Officially Released, Convenient for Building Vector Lake Architecture

Milvus introduces vector lake capabilities based on External Collection, supporting direct querying of vector data stored in DLF data lakes without repeated import into Milvus. This effectively reduces storage costs and achieves unified integration of data lake and vector retrieval capabilities.

Milvus — Milvus Manager Released

Milvus Manager is a visual management tool provided by Alibaba Cloud Vector Retrieval Service Milvus. Users can complete instance data object management, data operations, vector retrieval debugging, user and permission management, and runtime status viewing through the Web console. The tool also supports integration with Alibaba Cloud products such as DLF and RAM, further improving development, debugging, and operations efficiency.

Milvus — Single AZ Upgrade to Same-City Disaster Recovery Productized

Vector Retrieval Service Milvus now supports upgrading single availability zone instances to same-city disaster recovery instances, helping users enhance business high availability and disaster recovery capabilities, meeting scenarios with higher stability and continuity requirements.

Community