Embed 5 Pro scores 65.8 nDCG@10 on ViDoRe V3, second only to ColEmbed-3B-v2
qdrant_engine · x · 2026-10-23
On five public English ViDoRe V3 corpora, Cohere's single-vector Embed 5 Pro scored 65.8 nDCG@10, beating ColPali (54.8), ColQwen2.5 (61.2), Jina-v4 (65.2) and Nomic-7B (59.3). Only multi-vector ColEmbed-3B-v2 scored higher (67.5). Caveat: Embed 5's training data isn't published, so test set overlap is possible.
Related event: Qdrant Tests Cohere Embed 5: One Vector Per Page Beats ColPali at 64 Bytes(4 posts)→
More from Models
- 2-bit quantized DeepSeek 4.1 Flash tested locally: KDL degrades but still works — bakawolf123 · 2026-10-23
- KOL predicts K3.1 benchmarks sight unseen: benchmaxxed, below Sonnet in practice — teortaxesTex · 2026-10-23
- Sam Altman: OpenAI's Best Internal Model Won't Be Released Anytime Soon — rohanpaul_ai · 2026-10-23
- AI suddenly solves 80-year-old math problems: humans search breadth-first, LLMs go depth-first — Latent Space · 2026-10-23
- EmbeddingGemma 2 fine-tuned for 3 hours tops 8 Vietnamese ASR benchmarks at 356M params — stablequan · 2026-10-23
- Kimi K3.1 teasing drags on as users urge Moonshot to just ship it — scaling01 · 2026-10-23