On 2026-09-30 Cohere launched Embed 5 with Pro and Fast tiers that share one embedding space (index with Pro, query with Fast). Multimodal text/image/fused inputs, 128K context, 100+ languages; API ids embed-v5.0-pro / embed-v5.0-fast. Blog: ViDoRe V3 Pro avg 85.8 (+8.8 vs Embed 4); text pricing $0.12 / $0.08 per M tokens. Available on Cohere API, Model Vault, Azure Foundry, SageMaker; vLLM for private deploy.

Key Takeaways

  • ✓Shipped 2026-09-30: cohere.com/blog/embed-5 + docs.cohere.com/docs/cohere-embed
  • ✓Shared space: index Pro / query Fast; cross-model mean loss ~1.6–2.7% vs same-model on 40 dev sets
  • ✓ViDoRe V3: Pro 85.8 (+8.8 vs Embed 4) vs Voyage 4 Large 83.7 / Gemini Embedding 2 83.2; Fast 84.5
  • ✓128K ctx; dims 2048…256; float/int8/binary + Matryoshka; text $0.12/$0.08 per M (image $0.40/M both)
  • ✓API embed-v5.0-pro/fast; Foundry/SageMaker/vLLM; LangChain/Weaviate/Qdrant/Pinecone integrations
Cohere Embed 5 ships: Pro/Fast shared space; ViDoRe V3 Pro 85.8 avg; Fast at $0.08/M tokens
🖼️Official Media
Click to view high-res
🧭

Turn your technical choice into a development budget

Compare 40 dev plans & simulate token costs vs $20/mo subscriptions

🔬

In-Depth Technical Analysis

Core Background & Industry Pain Points

RAG/agents pay when retrieval is noisy. Enterprise corpora mix scanned PDFs, tables, slides, and many languages; running a “best” embedder and a “fast” one usually means two indexes. Cohere’s 2026-09-30 Embed 5 answers with Pro/Fast in one embedding space—index with Pro, query with Fast, no rebuild.

Architecture Highlights & Internals

Both tiers: text/image/fused inputs, 128K context, dims 2048…256, float/int8/binary + Matryoshka. API ids embed-v5.0-pro / embed-v5.0-fast. Shared-space cross tests (40 dev sets, Pro+Pro=100): Pro corpus + Fast query ~98.4. Storage example: 2048-d f32 ~8KB → 256-d binary ~32B; recommend 1024-d int8 for many deploys. Surfaces: Cohere API/Model Vault, Azure Foundry, SageMaker, vLLM private, batch embed for ingest.

Authoritative Benchmarks & Measured Scores

Vendor-reported (incl. RCP-nDCG@10). ViDoRe V3: Pro 85.8 (+8.8 vs Embed 4) vs Voyage 4 Large 83.7 / Gemini Embedding 2 83.2 / OpenAI text-embedding-3-large 75.5; Fast 84.5. FinanceBench 80.1/80.0; FinQA 90.0/88.8. Parsed-PDF suite Pro 84.8. Fast ~2.4× doc throughput vs Pro. Text price $0.12 / $0.08 per M; image $0.40/M both.

Developer Hands-on Guide

Follow https://cohere.com/blog/embed-5 and https://docs.cohere.com/docs/cohere-embed; index with Pro, query with Fast at matching dim/quant; validate on your corpus; deploy via API, Foundry, SageMaker, or vLLM.

Action HubWant to see how today's models and tools compare?

This story is not about one specific product. Compare leading models side by side, or see which coding plan fits your team.

Up NextSGLang-Neuron v0.1.0 Goes Open Source in Public Preview: SGLang Runs Natively on AWS Trainium, with Kimi-K3 Disaggregated Across Three Nodes and GLM-5.2 at 1M Context, Apache 2.0
Read Now→