Here's a Qwen 3.8 27b NVFP4 w/ dflash2 result, medium thinking, 32k reasoning budget, 11 minute run.
Not as intricate obv. but decent
Qwen 3.8 Flash Next has some wild numbers.
Here are the ones to remember: → 125B total parameters. → Only 6B active per token. → 51B engram embeddings. → 262K context out of the box. → Up to 1M tokens with YAN. → 62.5 on SWE-Bench Pro. → 73.9 on agentic office
Qwen 3.8-27B is not just another free AI model.
It can handle real SEO agency workflows. You can use it to: → Read huge documents. → Analyze coaching calls. → Watch videos. → Process images. → Write code. → Build automation plans. → Create structured reports. And
Your phone can't run the new Qwen 3.8 27B.
Not alone, anyway. I built SwarmLLM: it splits the model across the devices around you and runs it in browser tabs. Here it's 27B on a MacBook + an iPhone, peer to peer, no server. Built the peer-to-peer WebGPU inference engine from
You have reached the end of the archive
All of qwen38