Back to Arena

Cohere Embed

by Cohere Inc.

System Card

OrganizationCohere Inc.
Released2021-01
Architectureexternal-memory-network / multilingual multimodal embedding API
DetailsCohere Embed v3 (released November 2023) is a multimodal embedding model supporting 100+ languages and image+text encoding, achieving SOTA on MTEB at release. The model uses compression-aware training to produce high-quality int8 quantized embeddings. Embed 4 (2025) added further multimodal improvements at $0.12/MTok.
Parameters—
Domainrag-retrieval
Open SourceNo
WebsiteVisit
multilingualmultimodalint8-compressionenterpriseMTEB

Capability Profile

Benchmark Scores

5 of 14 benchmarks
Long-Context Retrieval
2/5
RULER
7582p
NIAH
no data
LooGLE
no data
LongBench
603p
∞Bench
no data
Multi-Turn Recall
0/2
LoCoMo
no data
MemoryBank
no data
Cross-Session Memory
0/1
LongMemEval
no data
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
70.943p
HotpotQA
7044p
Agent Task Memory
0/1
AgentBench-Mem
no data
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
1/1
RAGAS
6734p
Sources:Cohere Embed vendor documentation; evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)Cohere Embed vendor documentation; evaluated on LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding (Tsinghua KEG, 2308)Cohere Embed vendor documentation; evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)Cohere Embed vendor documentation; evaluated on RAGAS: Automated Evaluation of Retrieval-Augmented Generation (Exploding Gradients, 2309)Cohere Embed vendor documentation; evaluated on RULER: What's the Real Context Size of Your Long-Context Language Models (NVIDIA, 2404)