Quick question about how text-only/non-multimodal models are evaluated on OpenHands Index: as far as I know, DeepSeek-v3.2 and MiniMax M2.5 are both text-only models, so how does OpenHands Index score them on the Frontend task/SWE-Bench Multimodal? Does this mean those models don't take any image input at all and only take the text descriptions of the problem?

Yes!

Thanks!