Embeddings & Vector SearchMilvus (Zilliz)Community

Milvus 2.6 released

Released Milvus 2.6 — claims RaBitQ 1-bit quantization cuts memory by 72% and speeds up queries 4× (IVF_RABITQ compresses to 1/32 of the original), and introduces Woodpecker, a WAL built on object storage

Why it matters

The axis of vector DB competition moved from adding features to cutting the cost of billion-scale vectors

Source — Primary sources first; secondary reports are labelled “Press”

Officialmilvus.iohttps://milvus.io/blog/introduce-milvus-2-6-built-for-scale-designed-to-reduce-costs.md

Added in Week 35, 2026 · See that week

Same week

Naver AI center head named Korea's first Senior Secretary for AI Future PlanningPressAnthropic publishes the design of its multi-agent research systemOfficialCognition publishes Don't Build Multi-AgentsOfficialMistral AI announces Mistral ComputeOfficial

Nearby on the same axis

Alibaba Qwen presents Qwen3 EmbeddingPaperAlibaba Qwen releases the Qwen3 Embedding and Reranker series (0.6B–8B)OfficialLanceDB announces a $30 million Series A and launches its Multimodal LakehouseOfficialJina AI releases jina-embeddings-v4 (3.8B)Paper