Cohere releases Embed v4

Released Embed v4 — 128K context, unified embeddings for mixed image+text input, support for Matryoshka dimensions (256/512/1024/1536)

Why it matters

Commercial embedding competition moved beyond text search to long documents and mixed-modality search

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

Officialdocs.cohere.comhttps://docs.cohere.com/changelog/embed-multimodal-v4

Added in Week 35, 2026 · See that week

Same week

Gemma 3 QAT int4OfficialOpenAI launches o3 and o4-mini, with thinking with imagesOfficialPreparedness Framework v2OfficialCohere launches Embed 4Official

Nearby on the same axis

Cohere launches Embed 4OfficialCornell Tech researchers present Harnessing the Universal Geometry of EmbeddingsPaperGoogle DeepMind researchers present Gemini EmbeddingPaperLanceDB announces a $30 million Series A and launches its Multimodal LakehouseOfficial