ALIBABA recently announced comprehensive updates with a roadmap for its full-stack AI strategy to fully unleash the value of AI at scale to customers through its latest innovations.
At this year’s Apsara Conference, Alibaba Cloud’s annual flagship technology conference, the technology innovator announced a series of updates featuring Qwen foundation models and multimodal models, new proprietary AI chips, a purpose-built agentic cloud, and an AI agent platform for mobile phones, reinforcing Alibaba Cloud’s leading global position in the new AI era.
“The theme of this year’s Apsara Conference is ‘Intelligence Goes Beyond’. Over the past few years, AI has continuously expanded our imagination of technological capabilities. AI possesses vast potential for development–it can be deployed and scaled in real-world scenarios, boosting productivity across thousands of industries,” said Joe Tsai, chairman of Alibaba Group. “This is the true meaning of ‘Intelligence Goes Beyond’: guiding AI from technological breakthroughs toward value creation.”
“Today, the total volume of Machine Thinking is less than 3% of all Human Thinking. If that volume eventually scales to 1,000x human capacity, the simple math tells us: Machine Thinking still has an enormous growth runway. As machines are becoming the primary force behind Thinking, turning intelligence into a commodity supplied at scale, the truly groundbreaking products of the Machine Intelligence era have not yet arrived,” said Eddie Wu, CEO of Alibaba Group. “With this in mind, our target is that by 2032, the global data center capacity operated by Alibaba Cloud will surpass 20GW, fueling the industry’s exponentially rising demand for AI.”
Qwen 4 in training
At the conference, Alibaba revealed that its next-generation model, Qwen 4, is currently in training. The company further announced its roadmap for the upcoming Qwen 4.5 and Qwen 5 model series, projected to scale up to 5 to 10 trillion parameters.
Alibaba has revealed progress in RSI (Recursive Self-Improvement) driven by empirical feedback. Over a month of fully automated runs – spanning pipeline design, data validation, iterative experimentation, and error diagnosis – Qwen3.8-Max completed 33 iterative cycles. Through autonomous training optimisation and post-training techniques, the updated Qwen3.8-Max boosted its Artificial Analysis score from 40 to 45.
In a chip design experiment, the model underwent over 60 hours of self-improvement across the entire design lifecycle, making more than 10,000 EDA tool calls to produce production-grade chip bus modules. This reduced chip area by 42% with zero compromise in performance.
Multimodal Model Family Upgrades
Alibaba introduced multimodal model updates spanning speech, audio, and vision.
The company debuted Qwen3.8-LiveTranslate, a simultaneous interpretation model that enhances translation fidelity, fluency, and brevity, and reduces latency (LAAL) by nearly 20% from 2.8 to 2.3 seconds for a smoother, more responsive real-time translation experience.
It also unveiled Qwen-Audio-3.1-TTS-Next, a next-generation audio generation model capable of creating complete cinematic soundscapes by blending dialogue and ambient sounds all at once from a text script, designed for professional creative needs such as audiobooks, film and television, podcasts and games.
The updated speech suite also includes Qwen-Audio-3.1-ASR (Automatic Speech Recognition), Qwen-Audio-3.1-TTS (Text-to-Speech), and Qwen-Audio-3.1-Realtime, offering enhanced voice AI capabilities for understanding, generation, and real-time interaction.
Qwen-Image 3.1, an image-generation model optimized for creative design and e-commerce marketing, is set to launch later this year. It cuts visual design turnaround from hours to seconds while offering advanced generative tools, including native transparent-background generation and versatile image-editing capabilities.
To standardize world model evaluation, Alibaba Token Hub, alongside leading academic and industry partners, has introduced a six-tier capability framework and launched Happyworld Arena benchmarking platform, spanning video generation, embodied AI, and spatial reconstruction.
Bringing the models closer to general users, Alibaba has launched Qwen Intelligence, a business-facing full-stack agentic solution optimized for smartphones. It provides phone makers with access to a Qwen-powered agent platform to support next-generation AI phones that can handle complex, cross-app tasks reliably.
T-Head Unveils Zhenwu V900 and Yitian CPU Roadmap
T-Head, Alibaba’s chip design unit, has unveiled the Zhenwu V900, its latest AI training and inference processor featuring high-capacity memory and robust inter-chip bandwidth.
The Zhenwu V900 delivers three times the performance of its predecessor, the Zhenwu M890 (released in May). Featuring 216 GB of GPU memory and 1,200 GB/s of inter-chip bandwidth, the new accelerator is built to power demanding AI workloads with native support across multiple data precisions, including FP8 and FP4. By significantly reducing inference costs while increasing compute density, it seamlessly handles both high-precision model training and ultra-low-precision inference. It is scheduled for mass production and commercial release in Q1 2027.
T-Head’s Zhenwu AI chips have been serving over 650 customers across different industries, including automobile, finance, large language models, embodied intelligence, energy, and manufacturing.
Alibaba also unveiled its upgraded supernode server, which integrates the Zhenwu V900 processor, ICN Switch, Panmai SmartNIC, and Zhenyue SSD controller chip. Optimized for full-stack system-level synergy across compute, storage, and networking, the server can support a supernode cluster comprising up to 500,000 cards.
Additionally, Alibaba unveiled its roadmap for next-generation proprietary CPUs tailored for agentic AI tasks, scheduled for launch in 2027. The Yitian 720 features enhanced single-core performance, higher core density, and increased energy efficiency compared to its predecessor, the Yitian 710. The Yitian 730, the first CPU built on T-Head’s proprietary microarchitecture, boasts up to a 40% increase in SPECint2017/GHz performance over the Yitian 710’s.
Agentic Cloud
Alibaba Cloud unveiled a comprehensive suite of upgrades centered on its agentic cloud strategy, built around three core scenarios — model, harness and context. The upgrades span three corresponding layers: AI Native Cloud for model, powering large-scale model training and inference; Agent Native Cloud for harness, enabling enterprise-grade agent deployment, operation and security; and Context Engine for context, providing real-time data and long-term memory, helping customers create real-world value with AI at scale.
AI Native Cloud
- Alibaba Cloud’s Platform for AI (PAI) has integrated synergistic optimizations integrating inferencing, caching, sample replay, and training services. In real-world post-training of Qwen models, PAI completed state-of-the-art model training in just five days, advancing from pre-training toward Agentic RL.
- Cloud Parallel File Storage (CPFS) system is a new-generation system optimized to better support AI model training, delivering hundred-terabyte-per-second throughput and hundred-million input/output operations per second (IOPS), helping cut enterprise AI storage costs by 69%.
- HPN 8.0 Pro, Alibaba Cloud’s latest proprietary AI networking architecture, delivers 100 petabits of bandwidth with ultra-low latency. A single cluster can support over 130,000 network ports at 800G speeds, with built-in redundancy ensuring that neither optical transceiver nor link failures cause service interruptions, while reducing the impact of network upgrades and failures from 50% to 25% compared to the previous generation.
Agent Native Cloud
- AgentCore is a new enterprise platform to build, run, and manage AI agents throughout their entire lifecycle. It enables businesses to easily build and run their AI tools, safely control human-agent collaboration, and monitor performance for continuous improvements. With enterprise-grade security embedded at the operating layer, AgentCore provides organizations with a single, controlled foundation to deploy diverse AI agents and seamlessly integrate them into critical business systems at scale.
- The Agent Security Center delivers full-lifecycle security and compliance management for enterprise AI agent applications. The platform provides a multi-layered defense system powered by real-time threat detection, ensuring agents operate securely within controllable boundaries at scale.
Context Engine
- Agent Context is an enterprise-grade context data service that gives AI agents real-time context and long-term memory. By connecting a company’s documents, business systems, chat records, and multimodal data into a single foundation, it enables agents to remember past tasks, share knowledge across teams, and learn continuously. In knowledge-intensive scenarios such as customer service, AI coding, and data analytics, it cuts token usage by up to 67%.
- OpenLake has been upgraded into a unified, multi-modality data lakehouse, where a single copy of data—structured, semi-structured, unstructured, vector, and streaming formats—serves multiple compute engines for processing, search, analysis, and model training. Compared with traditional architectures, OpenLake reduces total costs by 38% and cuts query response times by 40%.
