How much does the Embed 5 Pro API cost per million tokens?

Embed 5 Pro costs $0.12 per 1M input tokens and $0.00 per 1M output tokens. Prices are tracked from 1 paid provider.

How much does a typical Embed 5 Pro request cost?

A typical request costs $0.0012 at the cheapest provider, Cohere. The request size is 10,000 input tokens and 1,000 output tokens.

How reliable are Embed 5 Pro API endpoints?

Yes, the 1 endpoint with uptime data in the last day was at 99% uptime or better, and the best uptime was 100.0%, by Cohere.

What is the context window of Embed 5 Pro, and how long can its output be?

The context window is 128K tokens (128,000). Maximum output is 115K tokens (115,200).

Can Embed 5 Pro read images, audio, video or PDFs?

Embed 5 Pro accepts text and image inputs. Audio, video and PDF inputs are not listed.

What are the best alternatives to Embed 5 Pro at a similar price?

Matched on model kind, blended price and release date: voyage-code-4 by VoyageAI by MongoDB at $0.09 per 1M tokens, voyage-multimodal-3.5 by VoyageAI by MongoDB at $0.09 per 1M tokens, and Titan Embeddings G1 - Text by Amazon at $0.075 per 1M tokens.

Who made Embed 5 Pro and when was it released?

Embed 5 Pro was made by Cohere and was first listed on OpenRouter on November 3, 2026.

Is Embed 5 Pro free to use?

No, the listed API price is $0.12 per 1M input tokens and $0.00 per 1M output tokens. Its openness is unknown.

Embed 5 Pro

Cohere

Announcement

Embed 5 Pro

Developer	Cohere
Listed	November 3, 2026
Context window	128K tokens
Max output	115K tokens
Input	Text, image
Output	Embeddings
Website	cohere.com

Part of a series on

Cohere

input: $0.12
output: $0.00
Context: 128Ktokens
Listed: 1w ago

Embed 5 Pro is Cohere's quality-optimized tier within Embed 5, a new family of embedding models the company built for enterprise retrieval. The launch announcement claims it achieves the highest average retrieval score of any model Cohere tested, with particular strength on financial datasets, parsed PDFs, and visually rich documents. It is built for search, RAG, and agentic retrieval, supplying relevant context to downstream generative models.