← Back to Model Beat
Models·Aug 27·all news from August 27, 2026

Alibaba Previews Qwen4 Architecture: 6B Active Params, One-Ninth the Training FLOPs

Alibaba has unveiled its Qwen4 architecture, which utilizes 6 billion active parameters. By requiring only one-ninth the training compute typically needed for models of this scale, the design suggests a potential shift toward greater efficiency in model development.

Covered by 1 source

Related stories

ModelsAlibaba Releases Smaller, Cost-Effective Qwen AI ModelAug 25 · 26 sourcesModelsIntelligent transcription with Gemini 3.5 TranscribeAug 26 · 7 sourcesModelsGLM-5.3-Flash matches top models at a fraction of the cost, and runs without NvidiaAug 26 · 11 sourcesModelsGemini Omni 1.1 Flash lets you build with more controlAug 27 · 4 sources