
Alibaba ($BABA) just dropped Qwen3.8-27B and it's literally matching Claude Opus 4.6
Alibaba’s AI team just sneak-dropped a nuke on the entire AI sector. They just released their new open-weight model, Qwen3.8-27B, on Hugging Face.
https://huggingface.co/Qwen/Qwen3.8-27B
This thing is only 27 Billion parameters. For the smoothbrains, that means it's incredibly tiny. But it is benchmarking right at the Claude 4.6 Opus Max level. It’s straight-up beating massive 100B+ models in agentic coding and complex reasoning. Anthropic and OpenAI are spending billions on compute to build massive closed models, and Alibaba just open-sourced a tiny beast that does the exact same job for free.
You can run this on a gaming rig. Because it’s so small, you don’t need a $40,000 Nvidia H100 to run it. This thing fits comfortably on a single 24GB VRAM GPU. you can easily run this on a 3090, 4090, 5090, or a Mac with 32GB+ Unified Memory.
TL;DR: Alibaba released a tiny AI model (Qwen3.8-27B) that punches like a heavyweight champion (matches Claude 4.6 Opus Max). It runs on consumer GPUs. Open-source is eating the AI software moat.