H company open-sources NeoMME: a multimodal-native encoder with no separate vision tower

NeoMME: an efficient Multimodal-native and Multilingual Encoder

H company released NeoMME, a family of 260M and 800M multilingual multimodal encoders. It uses a single bidirectional Transformer for both text tokens and raw image patches, trained from scratch with a masked discrete-diffusion objective—no separate vision tower, no causal LM. The fine-tuned NeoMME-Retriever outputs dense and late-interaction embeddings in one forward pass. Both sizes sit on the ViDoRe v3 Pareto frontier for nDCG@10 vs. model size. At 2048×2048 input on an L40S GPU, the 260M model encodes ~51 pages per second, roughly twice ColM's speed. The post does not disclose training data size or the full list of supported languages.