Been waiting for Qwen 4 all weekend.
But watching 54 t/s Qwen 3.8 Flash Next FP4 Local on a DGX Spark still so much fun
Low Polly House- Qwen 3.8 Flash Next threejs
Qwen 4 A couple 1 shot shaders.
I remember these used to take a couple tries with Qwen 3.8 max, most came out well however this time around haven't had a single failure.
I made AI models play badminton to find out which one can reason fast AND accurately 🏸
And I was shocked: OpenAI's own open-source model GPT-OSS-120B scored higher than GPT-6 Luna and GPT-5.6 Sol. How it works 👇 Each AI gets the game state as numbers: • the ball's x/y
You have reached the end of the archive
All of qwen38