Qwen 3.8 27B is now on Ollama.
That means you can run a serious multimodal AI agent on your own computer. No cloud login. No waiting. Offline after download. And its benchmark gains over Qwen 3.6 are hard to ignore.
Qwen-3.8-27b runs very usefully on my home PC -- no cloud servers required!
I've heard a lot people talk about the "AI bubble" but this is the actual realistic scenario for it: the hoi-polloi being able to run Opus-level models on their old gaming PCs. (RTX4070, i9, 32GB)
Qwen 3.8 27b running fully on my mac just recreated skribbl.
We’re entering the era of infinite local intelligence. 🤯
I didn’t know it before, until I used DGX Spark to test Qwen 3.8-27B and Ling-3.0-flash in the past few days.
I came to the conclusion that Spark’s memory bandwidth is a flaw😅 1. It is destined that large models can run, but they cannot run fast. 2. The small model can run, but the graphics card cannot run it. Screen recording: Qwen 3.8 27B multi-concurrency test, maximum 32 tok/s 🙃…
You have reached the end of the archive
All of qwen38