WeMM-JEV

27B multimodal decision model on Qwen3.8-27B taking text plus multiple images: choice / noul / score with a probability per option in one forward pass. #12 on the Nov 6 draw (68.49); its own public-panel run posts 77.97 on vision items against 69.37 on text, with the results dataset published. Apache 2.0, card on the thin side — the multimodal standing is the story.