LightOn Releases Fully Open Multilingual Retrieval Models mDenseOn & mLateOn
antoine_chaffin · x · 2026-07-30
LightOn has released mDenseOn and mLateOn, two fully open 307M-parameter multilingual retrieval models.
- Data & Training: They extended their validated English recipe to 8 additional languages via translate-train, creating a massive 2.8B-pair pre-training corpus and 16.3M fine-tuning samples covering natural languages and code.
- Performance: The models achieve SOTA on benchmarks like BEIR (English), MIRACL (multilingual), and MLDR (long-context). Notably, multilinguality does not degrade English performance, with mLateOn hitting 57.56 on BEIR.
- Generalization: The late-interaction architecture (mLateOn) shows superior zero-shot transfer to unseen languages, scoring 67.59 vs 57.42 for mDenseOn on the MIRACL unseen-language split.
Models, datasets, and training code are fully open-sourced.
More from Models
- Gemini Robotics Model Accurately Splits 9-Minute Excavator Video into 40+ Clips — DynamicWebPaige · 2026-07-30
- Kimi K3 State Loss After Compaction: A Deep Dive into API Payloads — Few_Sort8392 · 2026-07-30
- Report: Anthropic Opus 5 Incoming, Kimi K3 to Be Largest OSS Model — WolframRvnwlf · 2026-07-30
- DeepMind Launches Gemini Robotics 2 for Whole Body Intelligence — ai2027 · 2026-07-30
- Google Launches Gemini Robotics ER 2 for Embodied Reasoning — OfficialLoganK · 2026-07-30
- Anthropic Lacks Standard Completions Endpoint; Fair Model Eval Needs Unified Harness — altryne · 2026-07-30