Lighton Releases 307M Parameter Open Multilingual Retrieval Models
tomaarsen · x · 2026-07-30
Lighton has released mDenseOn and mLateOn, two open-source 307M-parameter multilingual retrieval models.
- Training Data: Trained on a 2.8 billion-pair translate-train corpus, one of the largest open multilingual retrieval training sets to date.
- Performance: mLateOn achieves the best results on BEIR, target-language MIRACL, and MLDR benchmarks, while remaining competitive on full MIRACL and code retrieval.
- Generalization: Experiments reveal mLateOn generalizes significantly better to unseen languages and scripts. On unseen MIRACL languages, it averages 67.59 compared to mDenseOn's 57.42.
- Open Source: Models, datasets, and training code are fully open-source.
Related event: LightOn Open-Sources Multilingual Retrieval Models mDenseOn and mLateOn(18 posts)→
More from Models
- Rumor: GLM 5.5 and Kimi K3.1 dropping soon, possibly slowed by anti-distillation blocks — zephyr_z9 · 2026-10-16
- Thrive Holdings benchmarks AI on real tax and IT work across 75+ firms — BorisMPower · 2026-10-16
- A Little Encouragement Makes a Small Local Model Draw a Stunning Pelican SVG — Fancy-Snow7 · 2026-10-16
- Rapid-MLX 0.16.2 brings NVIDIA Nemotron 3.5 up to 120B to Apple Silicon Macs — kuchaev · 2026-10-16
- "End of Chinese OSS" claim pushed back: open models get license almost automatically — eliebakouch · 2026-10-16
- Opus 5.5 still leads a month later, but open-weight decision models are closing in — airesearch12 · 2026-10-16