--- title: Cohere’s faster query model barely dents retrieval... description: Cohere released Embed 5, introducing a two-model split: the Pro model for indexing and the cheaper, faster Fast model for queries, sharing a single embedding... canonical: https://daily.dev/posts/cohere-s-faster-query-model-barely-dents-retrieval-quality-in-its-tests-6dwsqnozu twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Cohere’s faster query model barely dents retrieval quality in its tests | daily.dev og:description: Cohere released Embed 5, introducing a two-model split: the Pro model for indexing and the cheaper, faster Fast model for queries, sharing a single embedding... og:url: https://daily.dev/posts/cohere-s-faster-query-model-barely-dents-retrieval-quality-in-its-tests-6dwsqnozu og:image: https://api.daily.dev/og/posts/6DwsQNOZU.png og:image:alt: Cohere’s faster query model barely dents retrieval quality in its tests og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Cohere’s faster query model barely dents retrieval quality in its tests **[The New Stack](https://daily.dev/sources/newstack)** · 5 min read · 0 upvotes · 0 comments ## Summary Cohere released Embed 5, introducing a two-model split: the Pro model for indexing and the cheaper, faster Fast model for queries, sharing a single embedding space so teams avoid re-embedding their corpus. In Cohere's own tests across 40 datasets, Fast queries against a Pro index scored 98.4 versus a 100 Pro-to-Pro baseline, while Fast-to-Fast dropped to 96.6. Pro costs $0.12 per million tokens versus $0.08 for Fast, with Fast delivering roughly 2.4x document throughput. Both models support six vector dimensions (256-2,048) with float32, int8, and binary formats; Cohere recommends 1,024-dimensional int8 for most deployments. Embed 5 also supports text, images, and fused text-image inputs across 100+ languages with a 128K-token context window, and is evaluated with a new RCP-nDCG@10 metric that isn't directly comparable to prior scoring methods. The models are available via Cohere's API, Model Vault, Microsoft Foundry, Amazon SageMaker, and on-prem via vLLM. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Can I index with Cohere Embed 5 Pro and query with Embed 5 Fast without rebuilding my vector index? Yes, Embed 5 Pro and Embed 5 Fast share the same embedding space and produce compatible vectors at the same dimensions, so teams can index with Pro and query with Fast without re-embedding the corpus. Cohere's tests found this combination scores 98.4 versus a 100 baseline for using Pro on both ends, a small retrieval quality tradeoff for faster, cheaper queries. _daily.dev surfaces practical breakdowns like this for teams weighing RAG infrastructure tradeoffs._ ### What vector dimension and quantization does Cohere recommend for Embed 5 in production? Cohere recommends a 1,024-dimensional int8 vector for most deployments, which reduces memory and storage versus full float32 while keeping retrieval quality close to full precision. A 2,048-dimensional float32 vector for 100 million chunks takes about 819 GB, while the 1,024-dimensional int8 version cuts that to roughly 102 GB; a 256-dimensional binary vector shrinks it further to about 3.2 GB but with accuracy loss. _Engineers sizing vector storage for RAG pipelines can track these tradeoffs on daily.dev._ ### How much cheaper and faster is Cohere Embed 5 Fast compared to Embed 5 Pro? Embed 5 Fast costs $0.08 per million tokens versus $0.12 for Pro, and delivers about 2.4 times the document throughput in Cohere's own testing. This makes Fast suited to heavy query traffic in RAG and agent workloads, while Pro is recommended for indexing where retrieval quality matters more and volume is lower. _Teams comparing embedding model costs for RAG workloads can follow updates like this on daily.dev._ ## Similar posts on daily.dev - [Medium](https://daily.dev/posts/medium-xuslxsabv) · Medium · 0 upvotes · 0 comments - [Models & Pricing](https://daily.dev/posts/models-pricing-6vw4gyejf) · Hacker News · 2 upvotes · 0 comments - [Do You Need a Custom Embedding Model?](https://daily.dev/posts/do-you-need-a-custom-embedding-model--0goip7lmx) · DigitalOcean Community · 1 upvotes · 1 comments --- Tags: [#rag](https://daily.dev/tags/rag), [#vector-search](https://daily.dev/tags/vector-search), [#embeddings](https://daily.dev/tags/embeddings), [#cohere](https://daily.dev/tags/cohere) [View this post on daily.dev](https://daily.dev/posts/cohere-s-faster-query-model-barely-dents-retrieval-quality-in-its-tests-6dwsqnozu) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Cohere’s faster query model barely dents retrieval quality in its tests","url":"https://daily.dev/posts/cohere-s-faster-query-model-barely-dents-retrieval-quality-in-its-tests-6dwsqnozu","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/cohere-s-faster-query-model-barely-dents-retrieval-quality-in-its-tests-6dwsqnozu"},"datePublished":"2026-09-30T18:46:03.245Z","dateModified":"2026-09-30T18:46:31.753Z","description":"Cohere released Embed 5, introducing a two-model split: the Pro model for indexing and the cheaper, faster Fast model for queries, sharing a single embedding...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/31aa622b47dbc94cf5cf93b67034f90a?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/31aa622b47dbc94cf5cf93b67034f90a?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"The New Stack","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"The New Stack","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/newstack","url":"https://daily.dev/sources/newstack"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/cohere-s-faster-query-model-barely-dents-retrieval-quality-in-its-tests-6dwsqnozu","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"rag,vector-search,embeddings,cohere","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The New Stack","item":"https://daily.dev/sources/newstack"},{"@type":"ListItem","position":3,"name":"Cohere’s faster query model barely dents retrieval quality in its tests"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/cohere-s-faster-query-model-barely-dents-retrieval-quality-in-its-tests-6dwsqnozu#faq","mainEntity":[{"@type":"Question","name":"Can I index with Cohere Embed 5 Pro and query with Embed 5 Fast without rebuilding my vector index?","acceptedAnswer":{"@type":"Answer","text":"Yes, Embed 5 Pro and Embed 5 Fast share the same embedding space and produce compatible vectors at the same dimensions, so teams can index with Pro and query with Fast without re-embedding the corpus. Cohere's tests found this combination scores 98.4 versus a 100 baseline for using Pro on both ends, a small retrieval quality tradeoff for faster, cheaper queries. daily.dev surfaces practical breakdowns like this for teams weighing RAG infrastructure tradeoffs."}},{"@type":"Question","name":"What vector dimension and quantization does Cohere recommend for Embed 5 in production?","acceptedAnswer":{"@type":"Answer","text":"Cohere recommends a 1,024-dimensional int8 vector for most deployments, which reduces memory and storage versus full float32 while keeping retrieval quality close to full precision. A 2,048-dimensional float32 vector for 100 million chunks takes about 819 GB, while the 1,024-dimensional int8 version cuts that to roughly 102 GB; a 256-dimensional binary vector shrinks it further to about 3.2 GB but with accuracy loss. Engineers sizing vector storage for RAG pipelines can track these tradeoffs on daily.dev."}},{"@type":"Question","name":"How much cheaper and faster is Cohere Embed 5 Fast compared to Embed 5 Pro?","acceptedAnswer":{"@type":"Answer","text":"Embed 5 Fast costs $0.08 per million tokens versus $0.12 for Pro, and delivers about 2.4 times the document throughput in Cohere's own testing. This makes Fast suited to heavy query traffic in RAG and agent workloads, while Pro is recommended for indexing where retrieval quality matters more and volume is lower. Teams comparing embedding model costs for RAG workloads can follow updates like this on daily.dev."}}]} ```