Phancy Launches Phai-A1 Language Model and Phai-M1 Video Generation Model
Phai-A1 supports million-token context and local deployment; Phai-M1 improves the quality of generated live-action-style video.
BEIJING, CHINA/HONG KONG SAR – Media OutReach Newswire – 18 November 2026 – Phancy Group Co., Ltd. ("Phancy" or "the Group", stock code: 6682.HK), a leading full-stack enterprise-AI cloud services platform, announced the launch of two models developed in-house as part of its Phai series. Phai-A1 is a general-purpose language model optimized for efficient inference and long-context processing in applications such as software development and AI assistants. It supports a native one-million-token context window and can be deployed locally on a personal computer. Phai-M1 is a video generation model designed to improve how people appear and move in live-action-style footage, enhance visual detail and synchronize visuals with audio.Dr. Dai Wenyuan, Founder and Chief Executive Officer of Phancy, said, "The AI industry is moving from competition over model capabilities to a phase in which computing infrastructure, models and applications are developed together. The Phai series is an important step in strengthening our full-stack AI capabilities. By optimizing our in-house models and computing platforms together, we can connect the underlying infrastructure with business applications and make our models better suited to practical business needs. Feedback from real-world use will help us refine the models, strengthen our technology and make our products more competitive."
Phai-A1 Supports Efficient Inference and Local Deployment
Phai-A1 supports code generation, bug fixing, tool use and multistep tasks. In tests conducted under equivalent single-GPU conditions and using identical coding-agent tasks, Phai-A1 responded approximately 8 to 10 times as fast as Qwen 27B. For workloads that involve frequent model calls and require low latency, enterprises can run a lightweight model on demand to shorten response times and manage inference costs.
Phai-A1 performed well on software engineering benchmarks including SWE-bench Verified, SWE-bench Pro and SWE-bench Multilingual. On some evaluations, its results were comparable to those of models with more parameters. Phai-A1 natively supports a context window on the scale of one million tokens. Its post-training data covered very long texts. The model can process entire code repositories, long documents and large volumes of material, and support extended multi-turn conversations and complex tasks.
Developers and enterprises can also run Phai-A1 locally on a personal computer, depending on available hardware and data compliance requirements. This gives them more flexibility in how they deploy the model, manage data and control costs.
Phai-M1 Generates Synchronized Video and Audio for Live-Action Content
Phancy trained Phai-M1 with a focus on live-action film and television scenes. The model is designed to render facial expressions and body movements more naturally, improve visual detail and reduce common distortions in AI-generated video. It supports productions such as live-action series and AI-generated short-form dramas.
Phai-M1 generates visuals and synchronized audio in a single workflow. In testing, it generated a 15-second video at 720p resolution in about three minutes. For production teams and content studios that need a steady supply of footage, generation time is a factor in planning schedules and costs.
Computing Platform Optimization and Product Integration
With support from the ModelHub XC community, the Phai series completed a dedicated, end-to-end post-training optimization process tailored to the operating characteristics of Huawei's Ascend computing platform. The Phai series is integrated with Phancy's ModelHub XC, HAMi and ModelSphere products and with applications including PhanthyCode and PhanthyMovie. These integrations connect model development, computing infrastructure and AI applications.
Planned Release of Model Weights
Phancy plans to make the weights for Phai-A1 and Phai-M1 available to developers worldwide in the near future. The Group will continue to refine the models, work with technology partners and use developer feedback to guide updates. It will also expand use of the Phai series in software development, AI-assisted office work and video creation, and support private deployment in enterprise environments.
Users can try the Phai models at https://phai.phanthy.com.
Hashtag: #Phancy #PhaiA1 #PhaiM1
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 "Everyone for AI," Phancy Group is committed to bringing together more technologies, enterprises and developers to advance AI together.
