Cohere Embed v4 vs embed-english-v3.0
Between these two, Embed v4 is the better model for almost any new project: it reads 128K tokens instead of 512, handles mixed text-and-image documents like PDFs, and isn't limited to English. embed-english-v3.0 mainly makes sense if you already have an index built with it. One thing has changed, though: Cohere released Embed 5 on 30 September 2026, and its pricing page now lists only Embed 5, so if you're starting fresh, look there too.
Specs side by side
embed-v4.0 takes text, images, or mixed text and images, with a 128K-token context. You choose the output size: 256, 512, 1024 or 1536 dimensions (1536 is the default), so you can trade a little accuracy for smaller vectors.
embed-english-v3.0 is English only, takes text and images, caps each input at 512 tokens and always returns 1024 dimensions. Long documents have to be chunked to fit. Its Hugging Face repo holds only the tokenizer; the model itself runs through Cohere's API, SageMaker or a private deployment.
Vectors from different models aren't interchangeable, so moving from v3 to v4 (or to Embed 5) means re-embedding your whole corpus.
Price, and where Embed 5 fits
Cohere's pricing page lists Embed 5 Pro at $0.12 per million text tokens and Embed 5 Fast at $0.08, with images at $0.40 for both. It no longer shows prices for v4 or v3; check your dashboard or cloud console. On Amazon Bedrock, Cohere's Embed English is priced at $0.0001 per 1,000 tokens ($0.10 per million). Trial keys are free but rate limited and not allowed for production or commercial use.
Embed 5 keeps the 128K context and the text-and-image input, adds a 2048 option at the top of the dimension list, and covers 100+ languages. Cohere reports it beats Embed 4 on its document-retrieval benchmarks (for example 84.8 against 78.6 on parsed PDFs), but those are Cohere's own numbers. Pro and Fast share one embedding space, so you can index with Pro and query with Fast.
Cohere Embed v4: Pick Embed v4 if you need long inputs, PDFs with images, or non-English text and you're on a platform where Embed 5 isn't available yet.
embed-english-v3.0: Pick embed-english-v3.0 only if your existing index was built with it and re-embedding everything isn't worth it right now.
Which one should you try?
Cohere Embed v4
FreemiumBest for Production semantic search requiring multimodal support
embed-english-v3.0
FreemiumBest for English-only semantic search applications
Pricing, capabilities and platforms side by side. Prices are as of the "Pricing checked" date in the table.
Cohere Embed v4
Cohere Embed v4 transforms text, images, and mixed content into unified vector representations for semantic search, RAG, document clustering, and similarity matching. The model supports 1,536-dimensional embeddings with flexible compression via Matryoshka embeddings (256, 512, 1024, 1536 dimensions). Priced at $0.12/1M text tokens and $0.47/1M image tokens, it delivers multimodal capabilities competitive with text-only alternatives. The API supports batch processing up to 128,000 tokens per request with asymmetric search optimization. Limitation: incompatible with v3 embeddings; corpus re-embedding required for upgrades.
embed-english-v3.0
embed-english-v3.0 generates semantic embeddings from English text, producing 1,024-dimensional vectors suitable for retrieval-augmented generation, classification, clustering, and semantic search tasks. It achieves state-of-the-art performance on MTEB and BEIR benchmarks and was trained on approximately 1 billion English training pairs. The model supports batches of up to 96 inputs with 512 tokens maximum per input, and supports both text and image embedding. Pricing is $0.10 per million tokens. A notable limitation is that it requires explicit input_type specification to differentiate between search documents, queries, classification, and clustering tasks.
| Attribute | Cohere Embed v4 | embed-english-v3.0 |
|---|---|---|
| Pricing | Freemium | Freemium |
| Price | $0.12 per 1M text tokens; $0.47 per 1M image tokens | $0.10 per million tokens |
| Pricing checked | Not checked yet | Not checked yet |
| Free trial | 0 days | Not stated |
| Open source | No | No |
| Has API | Yes | Yes |
| Self-hosted option | No | No |
| Platforms | Cohere Platform, AWS Bedrock, Azure AI Foundry, Amazon SageMaker, GitHub Models | Cohere API, AWS SageMaker, Azure AI Foundry, OCI Generative AI |
| Languages | English and 100+ languages for text input; English for image input | English (primary); text-image multimodal support |
| Released | 2025-04-15 | 2024 |
| Pros |
|
|
| Cons |
|
|
Made up your mind?
Cohere Embed v4
FreemiumBest for Production semantic search requiring multimodal support
embed-english-v3.0
FreemiumBest for English-only semantic search applications
Specs come from each tool's own site and docs (how we test). AIDiveForge is editorially independent.