Huazhi Future unveils Lingyan Miaoyu multimodal model

// 1h agoMODEL RELEASE

Huazhi Future unveils Lingyan Miaoyu multimodal model

Huazhi Future, a subsidiary of MAAS, has introduced its Lingyan Miaoyu 9B multimodal large language model and released its V1.0 enterprise AI application white paper. Built on the proprietary LingYu Architecture with ScopeEdit semantic editing, the model unifies text, image, video, and audio processing across an enterprise ecosystem integrating Xingchen liquid-cooled computing centers, the Tianyu scheduling platform, and the ALRO model portal.

// ANALYSIS

Huazhi Future's launch signals that enterprise AI adoption is shifting away from raw parameter races toward tightly integrated, sovereign stacks that unite modular physical compute, heterogeneous scheduling, and controllable domain-specific agent pipelines.

–**Full-Stack Convergence**: Pairs containerized, liquid-cooled infrastructure (Xingchen cabins accommodating high-density B300 clusters) directly with distributed scheduling (Tianyu) and a unified model gateway (ALRO).
–**Practical 9B Footprint**: Adopts an edge- and enterprise-friendly parameter scale optimized for targeted deployment, lower inference latency, and manageable operational costs.
–**Controllable Semantic Editing**: ScopeEdit aims to solve the enterprise pain point of reliably updating knowledge bases and editing domain facts without destabilizing core model reasoning.
–**Task-Based Computing Paradigm**: Reflects the ongoing transition from renting raw GPU hours to orchestrating multi-agent, SLA-driven enterprise workflows.

// TAGS

maashuazhi-futurelingyan-miaoyullmmultimodalagentai-infrastructurelingyu-architecture

DISCOVERED

1h ago

2026-09-21

PUBLISHED

1h ago

2026-09-21

RELEVANCE

5/ 10

AUTHOR

Maase_official

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

BENCHMARK10m ago

Additional benchmark results for mid-size open-source LLMs released on the Medmarks v1.0 leaderboard.

The creators of the Medmarks v1.0 open benchmark suite and leaderboard for LLM medical capabilities have published new results for recent mid-size open-source models, including Gemma, Qwen, Muse, and Nemotron. Their evaluation found that the Gemma 4 31B model currently leads this specific size class in medical tasks.

OPEN SOURCE12m ago

OpenShorts is an open-source AI tool utilizing Gemini and local models to automate the creation and publishing of vertical short-form videos.

OpenShorts provides a self-hosted alternative to SaaS tools like Opus Clip and Descript, automating the entire video clipping pipeline. It transcribes audio with faster-whisper, identifies viral moments using Gemini, reframes footage to 9:16 with face tracking, and adds animated subtitles. The platform also includes a built-in MCP server, allowing AI agents to programmatically control the video generation and publishing workflow.

OPEN SOURCE12m ago

Mailflare delivers a serverless, self-hosted email client and inbox running entirely on Cloudflare's edge platform.

Mailflare is an open-source, serverless email client and inbox built to run directly on Cloudflare Workers, D1, and R2 without dedicated server infrastructure. By pairing Cloudflare Email Routing for message delivery with D1 for relational data and R2 for attachment storage, it enables custom-domain mail management at minimal cost. The web inbox includes personal and shared mailboxes, automated incoming routing rules, rich-text composing, snoozing, search folders, and delegated access, serving as a lightweight self-hosted alternative to per-seat platforms like Google Workspace or Fastmail.

Huazhi Future unveils Lingyan Miaoyu multimodal model