Qwen3-VL took the show but this is also massive news. I'll give Qwen3-Max a try and see how it compares to Qwen3-235B-A22B-2507 on Qwen chat, I love that they already have a feature on the website where you can compare all models side by side for each prompt.

It feels like the Qwen3 is slowing taking the lead here, work they keep pushing is amazing.

One thing I do not understand yet is how Qwen3-235B-A22B-2507 has higher scores than Qwen3 Max (preview) on artificialanalysis, it is supposed to perform better?