Skip to main content

Alibaba's Tongyi Models Dominate Hugging Face Rankings

Alibaba Tongyi Models Lead Global AI Rankings

On September 28, Hugging Face announced its latest model rankings with Alibaba's Tongyi making a remarkable showing. Seven of its models now occupy top-ten positions in the global open-source rankings, highlighted by the debut of Qwen3-Omni which claimed the #1 spot.

Image

Qwen3-Omni Sets New Industry Standards

The newly open-sourced Qwen3-Omni represents a significant leap forward in multimodal AI capabilities. This advanced model achieves 32 state-of-the-art (SOTA) benchmarks in audio and video processing while maintaining exceptional performance in text and image tasks - an industry first.

Key capabilities include:

  • Processing four data types: text, images, speech, and video
  • Human-like "listening, speaking, and writing" functions
  • Stable performance across both multimodal and single-modal tasks

Previously requiring multiple specialized models, complex instructions can now be handled by a single Qwen3-Omni instance. This breakthrough has major implications for deployment in:

  • Automotive systems
  • Smart glasses
  • Mobile applications

The consolidation of these capabilities into one model significantly enhances user experience while reducing implementation complexity.

Expanding Tongyi Model Ecosystem

The Qwen3-Omni joins six other high-performing Tongyi models that ranked in Hugging Face's top ten:

  1. Qwen3-VL: Advanced visual understanding
  2. Qwen-Image-Edit-2509: Sophisticated image editing
  3. Wan2.2-Animate: Action generation
  4. DeepResearch: Specialized research agent
  5. Two additional specialized models

The strong showing follows Alibaba's recent showcase at the 2025 Yunqi Conference where these models were introduced alongside other AI innovations.

Key Points:

  • Alibaba dominates with seven models in Hugging Face's top ten rankings
  • Qwen3-Omni debuts at #1 with unprecedented multimodal capabilities
  • Single-model solution replaces multiple specialized predecessors
  • Potential applications span automotive to consumer electronics sectors
  • Performance maintained across both multimodal and traditional AI tasks