Alibaba Announces Comprehensive Full-Stack AI Upgrade for the Agentic Era

Qwen3.7-Max, upgraded cloud infrastructure and model services, and new T-Head chips announced at Alibaba Cloud Summit

HANGZHOU, CHINA - Media OutReach Newswire - 20 May 2025 - Alibaba today announced a comprehensive upgrade of its full AI stack—spanning cloud infrastructure and model services, AI chips and foundation models —to empower customers in building, deploying, and scaling AI agents with greater efficiency, reliability, and performance.

Unveiled at the Alibaba Cloud Summit, Qwen3.7-Max is Alibaba's latest large language model, engineered for advanced agentic coding, complex reasoning, and long-horizon task execution. Qwen3.7-Max will be available soon for developers and enterprises worldwide.

To address surging compute and AI workload demands in the agentic era, Alibaba Cloud has also upgraded its infrastructure and model services. Key launches include the Panjiu AL128 Supernode Server, designed to empower scalable agent inference and large-scale model training, and an optimization update within Alibaba's model service platform that continuously refines model performance.

Additionally, T-Head, Alibaba's semiconductor design subsidiary, introduced the Zhenwu M890, its latest AI training and inference processor, featuring high-capacity memory, robust inter-chip bandwidth, and native FP4 precision support.

Qwen 3.7-Max: A Versatile Foundation Model for the Agent Era

Designed as a robust foundation for AI agents, Qwen 3.7-Max seamlessly handles code generation and debugging, office workflow automation, and complex multi-step tasks requiring hundreds or thousands of actions.

The model delivers exceptional agent capabilities across diverse domains. As a frontier-level coding assistant, it supports coding tasks from rapid frontend prototyping to complex, multi-file software engineering. To enhance office work productivity, it reliably orchestrates multi-agent workflows to tackle sophisticated operations. Notably, Qwen 3.7-Max can autonomously execute long-horizon agentic tasks—sustaining continuous operation for up to 35 hours and managing over 1,000 tool calls without performance degradation.

Deeply optimized for leading agent frameworks including OpenClaw, Hermes Agent, Claude Code, Qwen Paw and Qoder, it serves as a reliable backbone for different agent systems. The model achieves top-tier results across major benchmarks in coding, general-purpose agents, general capabilities and multilingualism, making it competitive with leading frontier models. It will be soon accessible through Alibaba's model service platform Model Studio for global developers.

Next-Generation Intelligent Computing and Enhanced Model Services

To empower scalable AI Agent inference and large-scale model training, Alibaba Cloud has launched the Panjiu AL128 Supernode Server, powered by the Zhenwu M890 AI processor and ICN Switch 1.0 networking chip. By tightly integrating 128 AI accelerators within a single rack, the system delivers single-rack bandwidth at the petabyte-per second (PB/s) scale, dramatically improving the handling of large-scale concurrent requests from agents.

The Panjiu AL128 is now available on Model Studio for the China market (or "Bailian"), enabling Chinese enterprises to efficiently address training and inference demands across sectors.

To optimize performance, Bailian has introduced Agentic RL, a reinforcement learning mechanism powered by agent execution feedback, to drive continuous model iteration. Bailian also features built-in safety governance capabilities, ensuring that autonomously operating agents always remain within defined boundaries.

T-Head's Latest Chips and Software Stack for AI Training and Inferencing

T-Head's latest AI accelerator, the Zhenwu M890, delivers three times the performance of its predecessor Zhenwu 810E. Zhenwu M890 features 144 gigabytes (GB) of GPU memory and 800 GB per second of inter-chip bandwidth. The chip natively supports multiple data precision formats, ranging from FP32 (32-bit floating-point) down to FP4 (4-bit floating-point), supporting both high-precision model training and ultra-low-precision model inference. These capabilities make it exceptionally well-suited for complex agentic AI workloads, which demand extensive working memory for context retention, high-speed communication for multi-agent coordination, and low-precision computing to maintain rapid execution while reducing cost. The chip is built on T-Head's proprietary parallel computing architecture and utilizes its custom ICN (Inter-Chip Network) interconnect protocol.

Alongside the accelerator, T-Head unveiled the ICN Switch 1.0, a dedicated switching chip designed to create high-bandwidth, low-latency scale-up networks for compute clusters. It delivers up to 25.6 Tbps of aggregate bandwidth and achieves extreme low latencyand congestion-free communication. By pairing the Zhenwu M890 with the ICN Switch 1.0 chip, it enables full-bandwidth interconnection across 64 accelerators, significantly boosting the computational efficiency and stability of large-scale intelligent computing. T-Head also unveiled its proprietary software stack, T-Head SAIL™, to unleash the full computational potential for its chips.

T-Head has achieved widespread industrial adoption of its proprietary AI chips, with over 560,000 Zhenwu units delivered to date. More than 400 external customers across 20 industries, including leading automakers and financial services companies, have deployed the chips to power intelligent operations.

Hashtag: #Alibaba

The issuer is solely responsible for the content of this announcement.

About Alibaba Group

Alibaba Group is a global technology company focused on e-commerce and cloud computing. We enable merchants, brands and retailers to market, sell and engage with consumers by providing digital and logistics infrastructure, efficiency tools and vast marketing reach. We empower enterprises with our leading cloud infrastructure, services and work collaboration capabilities to facilitate their digital transformation and grow their businesses.

** This press release is distributed by Media OutReach Newswire through automated distribution system, for which the client assumes full responsibility. **

Revenue Up 35.4% Year-on-Year API Token Call Volume Surges Nearly 6 Times

HONG KONG SAR - Media OutReach Newswire - 22 May 2026 - Phancy Group Co., Ltd. ("Phancy" or "The Company", stock code: 6682.HK), a leading general artificial intelligence company, today announced its unaudited consolidated results for the first quarter ended 31 March 2026.

During the period, Phancy achieved revenue of approximately RMB1.458 billion, representing a 35.4% year-on-year increase. Gross profit margin remained at 35.1%. Phancy leveraged its deep expertise in full-stack AI cloud services, to capitalize on the accelerating adoption of localized computing power and strong enterprise demand for AI solutions. The Company achieved robust growth in its core businesses, accelerated product innovation, and secured several major partnerships, sustaining strong operational momentum.

2026 First Quarter Business Highlights:

Unified Enterprise AI Platform Drives Explosive Core Business Growth

Global computing resources remain constrained, while demand for both private enterprise AI deployments and API-based model calls continues to grow rapidly. Phancy's enterprise-grade AI platform is built on a unified core architecture that seamlessly supports both API calling scenarios and dedicated private deployments. This significantly boosts AI application efficiency and resource utilization. Supported by a mature computing power supply chain developed over many years, Phancy's deployable computing power resources have increased by over 200%. This enables the Company to effectively meet surging Token demand and consistently deliver stable, high-quality AI services to its customers.

In the first quarter of 2026, API Token call volume surged nearly 6 times compared to the same period in 2025, and already accounted for nearly 40% of the full-year 2025 total. Meanwhile, the Agentic AI business expanded rapidly, with deepening commercial adoption. Orders on hand grew nearly 100% compared to the end of 2025, emerging as a major growth driver for the Company.

AI Technology Iteration Accelerates, Commercialization Beats Expectations

Building on its continued push into digital employee applications and AI empowerment across business units, Phancy has significantly shortened the product development cycle from R&D to commercialization, enhancing overall operational efficiency and customer satisfaction.

As of mid-May 2026, ModelHub XC has completed adaptation and optimization for over 70,000 AI models on domestic chips, achieving more than 70% of its full-year target - well ahead of schedule.

In May, Phancy launched PhanthyMovie, a professional-grade AI video generation platform designed to enhance creativity, control, and stability in video production, enabling standardized and large-scale content creation for the industry.

Leveraging its advanced technology and proven execution capabilities, PhanthyMovie achieved rapid commercial traction. Just days after launch, the Company entered into a strategic cooperation agreement with Huanxi Media, covering approximately US$200 million in AI Token usage. The two parties will also collaborate on the development of a next-generation AI-powered film and television content production platform, further strengthening Phancy's position in the AI-driven cultural and creative sector.

Core Products Align Closely with Policy Trends, Strengthening Compute-Model Integration

Since May 2026, China's AI sector has seen a series of positive policy developments focused on computing infrastructure, data element circulation, and open-source compliance governance. Phancy's core products, including HAMi vGPU and ModelHub XC, are well-aligned with national policy directions and mainstream industry trends.

In terms of computing resource allocation, policies emphasize cross-regional collaboration and broader access to computing power. Phancy's HAMi vGPU offers unified scheduling and fine-grained resource partitioning, effectively improving utilization rates, optimizing data center energy efficiency, and supporting unified management across multiple chips to boost single-card efficiency.

In data and model governance, the government continues to promote high-quality dataset development and compliance management. ModelHub XC supports multi-model adaptation and optimization, incorporates data traceability and security certification features to help enterprises reduce compliance risks, and uses the EngineX engine for batch adaptation of domestic chips and models. This significantly improves compatibility while enhancing Token output efficiency through targeted model tuning.

Through deep integration of its computing and model layers, Phancy has built a comprehensive "Compute–Model" integrated solution. This addresses key industry needs such as efficient computing utilization, secure data supply, enterprise compliance, and domestic substitution, while strengthening its technological moat. The Company is well positioned to capture policy dividends and industry opportunities, supporting enterprises in their digital and intelligent transformation.

Hashtag: #PhancyGroup

The issuer is solely responsible for the content of this announcement.

About Phancy Group

Phancy Group (6682.HK) is a leading full-stack AI cloud services platform, providing comprehensive solutions for the AI 2.0 era. Our offerings include SageAIOS, HAMi vGPU and ModelHub XC, delivering efficient and scalable AI infrastructure with end-to-end capabilities. We provide a complete solution from heterogeneous compute resource management and optimization to the deployment of intelligent agent models. These solutions empower digital transformation across a wide range of industries, supporting our vision of building a large-scale and efficient "Token Factory."

Guided by the mission of "AI for Everyone" and positioned as the "Navigator of AI," Phancy Group is committed to becoming a global leader in Artificial General Intelligence.

** This press release is distributed by Media OutReach Newswire through automated distribution system, for which the client assumes full responsibility. **