New Qwen AI Can Run on Your Own Computer
Run cutting-edge AI offline, free and private, on your laptop.
Deep Dive
llama.cpp’s new pre-release (b10660) adds support for Qwen3.8-Flash-Next (qwen4exp): new GGUF architecture, tensor loading, hparams, and a text graph with hyper-connections, gated delta net, MoE, and dense attention. It also includes the PLE n-gram hash embedding and was validated against vLLM.
Key Points
- Alibaba's Qwen team released a new experimental model, Qwen3.8-Flash-Next, designed to be fast and efficient.
- It's now supported by llama.cpp, a free program that runs AI chatbots entirely on your own hardware without internet.
- This is a developer preview — current users may face setup hurdles, but it signals a trend toward private, no-subscription AI.
Why It Matters
Private, free AI on your own device could end monthly subscriptions and stop your data from leaving your computer.