Alibaba’s new small model runs on a laptop and scores like a cloud one

Alibaba released the 27-billion-parameter Qwen3.8-27B model, which runs on a laptop with 17 GB of compressed weights and scores 52 on Artificial Analysis’s Intelligence Index, matching GPT-5.6 Luna. The model handles text, image, and video prompts, achieved 61.7 on SWE-bench Pro and 90.3 on LiveCodeBench, and surpassed Claude Opus 4.6 in those tests. It has been downloaded over three million times from Hugging Face within three days.