LightOn ships NeoMME: natively multimodal encoders at 260M/800M with no vision tower
antoine_chaffin · x · 2026-09-03
LightOn releases NeoMME, rejecting the common practice of using massive generative VLMs as representation models. Instead, the team trained encoders from scratch that are natively text-and-image, multilingual, and built for speed from the start — at just 260M and 800M parameters, with no vision tower.
- Also fine-tuned NeoMME-Retriever, a visual document retriever built on top
- The project delivers on Aurelien Lemonier's long-standing pitch of building a "multimodal ModernBERT"
More from Multimodal
- JSON prompting with MOSS Image 5B: bbox coordinates for exact multi-room layouts — cocktailpeanut · 2026-10-21
- MOSS Image 5B lands in Maestro with raw JSON prompting for precise layout control — cocktailpeanut · 2026-10-21
- FastVideo open-sources MiniMax FastH3 Omni with training code, 12x faster on Blackwell — multimodalart · 2026-10-21
- Megalith 008: stunning AI-generated colossal architecture series — azed_ai · 2026-10-21
- Higgsfield offers $25K prizes for open-source ComfyUI rebuilds of its apps — MuziqueComfyUI · 2026-10-21
- Stunning AI-generated elderly portrait blurs the line with real photography — LudovicCreator · 2026-10-21