Qdrant embeds 32.8M passages in 35 minutes on a laptop CPU with static embeddings
qdrant_engine · x · 2026-11-18
Qdrant and Minish detail how static embedding models (potion series, model2vec) — simple token-vector lookup tables with mean pooling, no transformer layers — let Fastembed embed 32.9 million passages in 35 minutes on a laptop CPU, versus 53 days for a transformer model. The article covers how static embeddings work, the optimizations added to Fastembed, and the retrieval-quality tradeoff of skipping inter-token interactions.
More from Infra
- BitDecoding-FP4: native FP4 KV cache decoding via algorithm-system co-design — UoE-Mlsys · 2026-11-18
- Lisa Su: Chip demand looks 'generational', not cyclical this time — BenBajarin · 2026-11-18
- Qdrant Generates 32.8M Embeddings in 35 Minutes With Static Embedding Models — qdrant_engine · 2026-11-18
- ComfyUI makes the MiniMax H3 video VAE up to 2.2x faster on Nvidia GPUs — NerdyRodent · 2026-11-17
- ArisGlobal cut Bedrock inference cost bloat by caching shared prompt prefixes first — AWS ML Blog · 2026-11-17
- Cloudflare researcher: Clef's fast release was 'asymmetric effort,' reasoning models next — michellechen · 2026-11-17