Phancy Launches Phai-A1 Language Model and Phai-M1 Video Generation Model
Phancy Group Co., Ltd. launched two in-house developed AI models on 18 November 2026: Phai-A1, a general-purpose language model, and Phai-M1, a video generation model. The company stated Phai-A1 supports a one-million-token context window and local deployment, while Phai-M1 improves live-action-style video quality and synchronizes visuals with audio.
On 18 November 2026, Phancy Group Co., Ltd., an enterprise-AI cloud services platform, announced the release of its Phai-A1 language model and Phai-M1 video generation model. Dr. Dai Wenyuan, Phancy's Founder and CEO, framed this launch as part of an industry shift, noting the AI sector is moving from competition over model capabilities to a phase integrating computing infrastructure, models, and applications.
The Phai-A1 language model is 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, allowing it to process entire code repositories, long documents, and large volumes of material. Phancy reported that in tests under equivalent single-GPU conditions and identical coding-agent tasks, Phai-A1 responded approximately 8 to 10 times as fast as Qwen 27B. Its performance on software engineering benchmarks, including SWE-bench Verified, SWE-bench Pro, and SWE-bench Multilingual, was described as comparable to models with more parameters.
For developers and enterprises, Phai-A1's ability to be deployed locally on a personal computer offers flexibility in deployment, data management, and cost control, depending on available hardware and data compliance requirements. For workloads requiring frequent model calls and low latency, running a lightweight model on demand could shorten response times and manage inference costs.
The Phai-M1 video generation model was trained with a focus on live-action film and television scenes. It aims to render facial expressions and body movements more naturally, enhance visual detail, and reduce common distortions in AI-generated video. The model generates visuals and synchronized audio in a single workflow. During testing, Phancy stated it generated a 15-second video at 720p resolution in about three minutes, a factor for production teams planning schedules and costs.
Both Phai series models are integrated with Phancy's ModelHub XC, HAMi, and ModelSphere products, as well as applications like PhanthyCode and PhanthyMovie. This integration connects model development, computing infrastructure, and AI applications. The models were optimized for Huawei’s Ascend computing platform with support from the ModelHub XC community. Phancy plans to make the weights for Phai-A1 and Phai-M1 available to developers worldwide in the near future.
Share this article
Related reading
6 stories
Anthropic blows past $2 trillion in Nasdaq debut, tops SpaceX’s June listing
We saw Anthropic's Nasdaq debut blow past $2 trillion, showing the market's appetite for new AI models.
Tennr, OpenAI to test AI agents for prior authorization
Tennr and OpenAI are already testing AI agents for prior authorization, a practical application of new models.

China’s battery-powered heavy trucks see electric growth

YeePay Adds 5 Partners to Expand Cross-Border Payment Network

Vietnamese Travellers Can Now Scan NETS QR Codes to Pay in Singapore

