Homi

About

Homi is an open model that takes an audio clip and a typed question and returns a probability for every answer option instead of generating text. It builds on Qwen3-Omni-30B-A3B-Instruct with frozen weights, a decision head and LoRA adapters, and supports yes/no, multiple-choice and ordered-score questions. The README reports 83.1 on MMSU against 85.7 for Gemini Flash, at 0.28 dollars per 10,000 questions against 14.80.