Tiny MoE ColBERT retrievers: 1.5M encoder to 115M model rivaling larger dense models
tomaarsen · x · 2026-10-26
VAGOsolutions released the SauerkrautLM-MoEdernColBERT collection on Hugging Face: multilingual ColBERT retrievers built on the MoEdernBERT architecture, available in 1.5M, 27M and 115M sizes with GGUF and CoreML builds.
Sentence transformers maintainer tomaarsen notes this scale of MoE retrieval is fascinating: the 1.5M encoder is tiny, while the 115M model competes with much larger dense models. He attributes part of the quality to multi-vector retrieval — each token keeps a 128-d vector and MaxSim lets query tokens match different document tokens, so far more vectors must be stored and scored than one dense vector per text.
Related event: VAGOsolutions releases tiny 1.5M-115M MoE ColBERT retrieval models(10 posts)→
More from Models
- Gemma 4 26B-A4B runs offline on a 16 GB Mac: 33/40 on TypeScript tasks at 31 tok/s — Mission-Sir-1614 · 2026-10-26
- Claude safety classifier blocks SS7 attack analysis, dev pivots to abliterated GLM — l4rz · 2026-10-26
- Sakana AI to pretrain first domestic model Sakana Sawara for defense and cybersecurity by 2027 — tkasasagi · 2026-10-26
- 31M-param English TTS model phonon-31m-en trends on Hugging Face — gradium · 2026-10-26
- NVIDIA tutorial: post-training Nemotron for Europe with Persona datasets and LoRA — NVIDIA Developer · 2026-10-26
- Opus 5 hailed as the 'most emo, greatest artist' model by AI community users — repligate · 2026-10-26