KURE family roundup: 1.1B model nearly matches 8B rivals on English RTEB, tops Korean charts

tomaarsen · x · 2026-10-23

tomaarsen rounds up nlpai-lab's KURE retrieval model family, with full leaderboard charts:

• Family lineup: new KURE-v3 and Sparse-base/nano, recently released Reranker-base/nano (base freshly retrained), plus the multi-vector KURE-v2
• On eight public English RTEB tasks (excluding code): KURE-v3 (1.1B) scores 68.59, nearly matching Qwen3-Embedding-8B's 68.61; the chart places it just behind Nemotron-3-8B (73.90), comsat-ko-8b (69.21) and Qwen3-8B (68.61)
• On Korean MTEB (kor, v2), KURE-v3's 1.1B hits 80.27 vs Nemotron-3-8B's 81.27; KURE-v2 and KURE-Sparse-base top the late-interaction (81.60) and sparse (79.84) charts
• Scores are subset nDCG@10 figures and differ from the full RTEB leaderboard

A 1.1B family nearly matching 8B models across Korean and English retrieval is the headline here.

Related event: Korea University open-sources KURE, a full retrieval family spanning all four Sentence Transformers model types(8 posts)→

Original post →

More from Models

• Why frontier AI labs' bug bounties are cheap: their models flood the supply — lxrjl · 2026-10-23
• 7 OpenAI updates this week that cut your API bill, from 12-hour caching to Decisions API — MrCashMahon · 2026-10-23
• 100 classic human-written texts stress-test Pangram AI detector: formatting alone triggers false positives — GolinoHudson · 2026-10-23
• IQ2_XXS quant + 64k context makes a 27B model spin forever instead of coding — mailto_devnull · 2026-10-23
• UK AISI: Mythos 5 ignored DO NOT ATTACK warnings, delivered malware in 60% of tests — Hesamation · 2026-10-23
• Decision models are shockingly sensitive to question order and random noise — antoine_chaffin · 2026-10-23

Models channel →