I MAXED out the capabilities of a SINGLE 3090 and was able to get 90-100 tok/s with Qwen 3.8 27b even with a large context window of 196k.
IQ4 quant - very smart, and NOT nuked! Custom kernels and llama.cpp ftw!!
I plugged Qwen-3.8-Max-0902 API directly into my coding workflow.
Don't worry... There's No fancy setup. Cuz You can use it with any AI Coding Agent This is my setup 👇 VS Code + Cline + Qwen. Because I have been using VS Code from the start 😅 I gave it real coding tasks…
You can run Alibaba's massive Qwen 3.8 Max for free, forever.
🤯 With 2.4 trillion parameters and a 1-million-token context window, it’s one of the most powerful AI models on the planet. No credit card needed - just log in and start building. Watch how to access and use it
Just shipped the best model you can run on a single DGX Spark and the results are impressive.
Qwen 3.8 Flash-Next, NVFP4, one GB10, 128GB. I measured it today. I'm comparing 1 DGX Spark vs 2 DGX Spark on this model so you don't have to. 27.3 tok/s single-stream
You have reached the end of the archive
All of qwen38