Alibaba on SoloSoft

AgentScope: Alibaba's Open-Source Multi-Agent Framework for Transparent AI Agents

Mon, 01 Jan 0001 00:00:00 +0000

Building production-grade multi-agent systems is notoriously complex. Coordinating communication between agents, managing distributed deployments, integrating with external tools, and ensuring observability are challenges that most frameworks tackle only partially. AgentScope, developed by Alibaba’s Tongyi Lab, addresses these challenges with a comprehensive framework designed for real-world, scalable multi-agent applications.

AgentScope distinguishes itself through its focus on transparency and controllability. Every agent’s decision-making process is observable, every message can be inspected, and the entire system can be configured through declarative specifications rather than imperative code. This makes it suitable for enterprise applications where auditability and reliability are paramount.

The framework supports both the Model Context Protocol (MCP) and Google’s Agent-to-Agent (A2A) protocol, enabling interoperability with a wide ecosystem of tools and agent platforms. Combined with its distributed communication system (MsgHub), AgentScope can orchestrate agent swarms that span multiple servers and geographic regions.

Alibaba Integrates Qwen AI with Taobao, Launches Agentic Shopping Experience

Mon, 01 Jan 0001 00:00:00 +0000

Why is Alibaba Choosing This Moment to Fully Embrace Agentic Shopping?

Direct Answer: Alibaba chose to fully integrate Qwen AI with Taobao in the second quarter of 2026 to launch agentic shopping features, aiming to differentiate through AI-driven experiences and widen the gap with competitors amid slowing growth in China’s e-commerce market. This transformation is not just about technological upgrades but a fundamental restructuring of the business model.

China’s e-commerce market has long entered an era of stock competition. According to data from the National Bureau of Statistics of China, the year-on-year growth rate of online retail sales in China dropped to 6.2% in 2025, far below the 16.5% in 2019. In such an environment, simple price wars or subsidy battles can no longer effectively drive growth. Alibaba chose to start from the fundamental aspect of “shopping experience,” completely rewriting the traditional “search-browse-order” process into a new paradigm of “dialogue-recommendation-automated execution.”

Animate Anyone: AI-Powered Character Animation from Single Images

Mon, 01 Jan 0001 00:00:00 +0000

Animate Anyone is a research project from Alibaba’s HumanAIGC group that turns a single photo into a fully animated video of a person walking, dancing, or performing any pose sequence – all while preserving the character’s identity, clothing, and appearance with remarkable fidelity. It represents one of the most impressive applications of image-to-video synthesis using diffusion models.

The core technical challenge Animate Anyone solves is temporal consistency with identity preservation. Previous approaches to character animation from single images suffered from flickering, appearance drift, and loss of fine details like clothing patterns or facial features. Animate Anyone’s innovation is a reference-guided diffusion architecture that injects appearance features from the input image into every frame of the generated video at multiple scales.

CosyVoice: Alibaba's Open-Source Multi-Lingual Voice Generation Model

Mon, 01 Jan 0001 00:00:00 +0000

Voice generation technology has seen remarkable progress, but most open-source text-to-speech (TTS) models still struggle with a fundamental trade-off: quality versus language coverage. CosyVoice, developed by Alibaba’s FunAudioLLM team, breaks this barrier by delivering production-quality voice generation across 9 languages and 18+ Chinese dialects.

With over 20,000 GitHub stars, CosyVoice has become a go-to solution for developers and researchers who need multilingual speech synthesis with advanced capabilities like zero-shot voice cloning, emotion control, and instruction-following generation. Unlike commercial TTS APIs that charge per character and limit customization, CosyVoice is fully open-source and self-hostable.

The model’s architecture is based on a novel approach that separates content, speaker, and style information into distinct latent spaces, enabling unprecedented control over generated speech. This design allows users to mix and match voices, languages, and speaking styles in ways that previously required extensive fine-tuning or separate models.

Higress: Alibaba's Cloud-Native AI Gateway Built on Istio and Envoy

Mon, 01 Jan 0001 00:00:00 +0000

As AI applications move from prototypes to production, the infrastructure layer for managing LLM API traffic has become critical. Organizations need to route requests to the right model, control costs with token-level rate limiting, cache responses intelligently, and monitor usage across teams and applications. Higress addresses all of these needs as a cloud-native AI gateway built on the battle-tested Istio and Envoy foundations.

Developed by Alibaba, Higress extends the traditional API gateway concept with native AI capabilities. It understands LLM request semantics – tokens, models, streaming responses, and prompt structures – enabling intelligent traffic management that goes far beyond what generic API gateways can provide.

Marco-o1: Alibaba's Open-Source Large Reasoning Model for Real-World Solutions

Mon, 01 Jan 0001 00:00:00 +0000

The race to build machines that can reason – not just pattern-match – has defined the cutting edge of artificial intelligence since the emergence of large language models. While proprietary systems like OpenAI’s o1-series have demonstrated impressive reasoning chains, the open-source community has long awaited a comparable alternative. Enter Marco-o1: an open-source large reasoning model from Alibaba’s AIDC-AI MarcoPolo Team that delivers structured, multi-step reasoning for both closed-form and open-ended problems.

Built on the Qwen2-7B-Instruct foundation, Marco-o1 represents a deliberate departure from models optimized solely for standardized benchmarks. The team at AIDC-AI designed it to tackle the messy, ambiguous problems that characterize real-world deployment – from logistics optimization to creative planning – while keeping the model fully open-source and accessible to the global research community.

MNN: Alibaba's Blazing-Fast Lightweight Inference Engine for Mobile and Edge AI

Mon, 01 Jan 0001 00:00:00 +0000

Running deep learning models on mobile and edge devices presents unique challenges: limited compute power, constrained memory, battery sensitivity, and diverse hardware architectures. MNN (Mobile Neural Network) is Alibaba’s answer to these challenges, a lightweight inference engine that brings AI to the edge with minimal overhead and maximum performance.

MNN powers over 30 of Alibaba’s applications, including Taobao (e-commerce), Youku (video streaming), and various enterprise tools. It has been battle-tested at billion-user scale, handling everything from real-time computer vision to on-device large language models. The engine’s small binary size (under 500 KB for the core runtime) and minimal runtime memory footprint make it suitable even for low-end devices.

Pixelle-MCP: Open-Source Multimodal AIGC Solution Bridging ComfyUI and LLMs via MCP

Mon, 01 Jan 0001 00:00:00 +0000

The Model Context Protocol (MCP) is reshaping how AI applications communicate, but most MCP tools remain narrowly focused on text and data queries. Pixelle-MCP shatters that limitation by turning ComfyUI – the most popular visual workflow engine for AI-generated content – into a full multimodal MCP server. Developed by Alibaba’s AIDC-AI team, this open-source solution lets any MCP-compatible client invoke complex AIGC pipelines for images, sound, video, and text using natural language.

The core insight behind Pixelle-MCP is elegant: instead of building multimodal generation capabilities from scratch, it repurposes ComfyUI’s vast ecosystem of community-built workflows as MCP-callable tools. Anyone who has designed a ComfyUI pipeline for stable diffusion, audio generation, or video synthesis can now expose that workflow to any LLM client as a simple API, with zero additional code.

Qwen Code: Alibaba's Open-Source AI Agent for the Terminal

Mon, 01 Jan 0001 00:00:00 +0000

Qwen Code is an open-source AI-powered terminal agent developed by the QwenLM team at Alibaba Cloud. Built from the ground up for the terminal environment, Qwen Code provides a Claude Code-style interactive coding experience optimized for Alibaba’s Qwen model family, while maintaining compatibility with models from OpenAI, Anthropic, Google, and others through a multi-protocol provider system.

The agent is designed to feel like a natural extension of the developer’s terminal workflow. It operates in the same shell environment, has access to the file system, can execute commands, edit files, create projects, and manage git workflows – all through natural language interaction. With support for agentic workflows that decompose complex tasks, sub-agents for parallel execution, and IDE integration via VS Code and JetBrains, Qwen Code positions itself as a versatile open-source alternative to proprietary coding assistants.

Qwen2.5-Omni: Alibaba's End-to-End Multimodal AI Model

Mon, 01 Jan 0001 00:00:00 +0000

Qwen2.5-Omni is Alibaba’s flagship open-source multimodal AI model, developed by the QwenLM team at Alibaba Cloud. As a single end-to-end model, Qwen2.5-Omni can perceive and understand text, images, audio, and video inputs simultaneously, while generating both streaming text and natural speech output – all within a unified architecture.

The model introduces several architectural innovations, most notably the Thinker-Talker architecture, which separates reasoning from speech generation while maintaining tight coupling between the two. With the introduction of TMRoPE (Time-Synchronized Multimodal Rotary Position Embedding), Qwen2.5-Omni achieves precise time alignment across modalities, enabling tasks like real-time video captioning, audio-visual question answering, and simultaneous interpretation.

VACE: Alibaba's All-in-One Video Creation and Editing Model (ICCV 2025)

Mon, 01 Jan 0001 00:00:00 +0000

Video generation and editing have traditionally been handled by separate models – one model for text-to-video, another for video stylization, yet another for inpainting. This fragmentation makes it difficult to build comprehensive video production pipelines and forces practitioners to learn multiple model interfaces. VACE (Video All-to-All Creation and Editing) eliminates this problem by unifying all video creation and editing tasks in a single diffusion transformer model.

Accepted at ICCV 2025, VACE is the work of Alibaba’s Tongyi Lab. The key insight behind VACE is that video creation and editing tasks share a common underlying structure: they all involve generating or modifying video content based on some combination of reference frames, text descriptions, and mask information. By designing a unified conditioning mechanism, VACE can handle all of these tasks without task-specific model variants.