Qwen 3.8 Max ran a project for 10 days without a human, spent 125 hours on a research paper
Wrote 7,600 lines of code and beat the original paper's result by 2.71 points while finishing ahead of 458 human teams in a separate ML competition.
Have you bought compute yet, anon?
Edited locally with the help of Qwen 3.8 27b
Name this mountain. Easter egg in the clip...
Built with Qwen 3.8 Flash Next (MiaAI_lab newest one Spark recipe).
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!!
You have reached the end of the archive
All of qwen38