20 hours later and I’m done!
Qwen 3.8 27B @ 73 TPS (peak) on a Macbook pro M5 max. MTPLX V2.7 out now with 3 new models. - Bare Speed: short burst decode, good for chat. - Optimized Speed: higher quality and faster on long coding. - Optimized Quality: 30% slower but super!
Yep, it did it, one prompt 'ish, first created the game, 2nd fixed a hang and made player smaller
3rd was play the game and make a video, it created a bot, played and recorded via playwright, 4th convert the produced webm file to mp4, QWEN 3.8 is exceptional.
Qwen 3.8 27B is still surprising me.
This is a Dotnet WPF App. It looks amazing.
Qwen 3.8:27B, with Ollama
On Nvidia 1660S & 32GB RAM. 👀
You have reached the end of the archive
All of qwen38