Alibaba's Qwen Overtakes Western Models in Open-Source Downloads as AI Power Shifts

Alibaba has emerged as the dominant force in open-source artificial intelligence, with its Qwen model family surpassing 1 billion cumulative downloads and spawning more than 200,000 derivative models to become the world's most popular open-source AI system, according to Time's ranking of influential AI companies.

The Chinese tech giant's appeal extends well beyond its home market. Airbnb has said it relies heavily on Qwen for its AI customer-service agent, citing quality and low cost, while Pinterest uses Qwen to analyze visual content and generate contextual text for pins. Alibaba is now attempting to convert its open-model lead into a full-stack AI empire, scaling cloud infrastructure, manufacturing proprietary AI chips, and selling agentic hosted versions of its models to enterprise clients.

The shift comes as many leading U.S. AI companies have retreated from open-source development, creating an opening that Chinese and European firms have moved to fill. Paris-based Mistral, founded three years ago and now valued at nearly $14 billion, has made openness a core business strategy. The company builds models for coding, transcription, document recognition, and multimodal tasks, helping customers run them on their own infrastructure—an appealing option for organizations that don't want sensitive data leaving the building.

"What matters is for customers to take AI systems and make them their own," Mistral CEO Arthur Mensch said. The strategy is paying off: Mistral's annualized revenue hit $400 million in early 2026, up roughly twentyfold from the prior year, driven by more than 100 major enterprise clients including ASML.

New York-based Hugging Face is positioning itself as a counterweight to concentrated AI power, operating the largest public repository for AI models, datasets, and applications—more than 2 million models and 500,000 datasets at last count. More than 30 percent of Fortune 500 companies have accounts on the platform. "We're staying on our mission because we think it's more important than ever," CEO Clément Delangue said. The company has pushed aggressively into AI agents and robotics, launching an open-source robot called Reachy Mini and a smolagents framework for building tool-using assistants with minimal code.

DeepSeek, whose low-cost AI model stunned the industry, launched a preview of its V4 model adapted for Huawei chip technology, underlining China's growing autonomy in the AI sector. The company's models have consistently been among the most used on international platforms hosting open-source models, despite Western governments banning their institutions from using DeepSeek over data privacy concerns.

(The U.S. State Department issued a global warning about alleged AI intellectual property theft by DeepSeek and other Chinese firms, though the Chinese Embassy in Washington has called the accusations groundless and deliberate attacks on China's AI industry progress. OpenAI has warned U.S. lawmakers that DeepSeek was targeting ChatGPT and other leading AI companies to replicate models for its own training, Reuters reported in February.)

The competitive dynamics reflect a fundamental split in AI development philosophy. While U.S. firms increasingly favor proprietary systems with higher inference costs, Chinese companies are betting on cost leadership and accessibility. DeepSeek's V4 Flash could become a default for cost-sensitive production use, while its V4-Pro-Max targets high-end applications. The company's 1 million context standard and open-sourcing approach allow customization and fine-tuning by developers worldwide, potentially shifting market dynamics toward accessible long-context AI.

The open-source momentum extends beyond model development into infrastructure and tooling. Brian Gerkey, CTO of Intrinsic, is scheduled to give an overview of the current state of open source in robotics and artificial intelligence at the Robotics Summit & Expo. The event will feature more than 50 sessions on AI integration into commercial robotics, with speakers from Tesla, Toyota Research Institute, and other major players exploring how perception, data integration, and continuous learning are redefining autonomy in real-world environments.