Imagine searching billions of images, documents, or products and finding exactly what you need in milliseconds. **Similarity search makes this possible, transforming how we interact with massive, complex datasets.**

It doesn't rely on matching exact words. Instead, it understands the deeper meaning behind your query.

This technology translates abstract concepts into mathematical representations. Computers can then instantly compare and retrieve these representations. From recommendation engines to medical research, this technique is reshaping how machines understand and navigate information with unprecedented precision and speed.

## Understanding similarity search

When you search for "black leather boots" on an e-commerce website or look for "songs that sound like Taylor Swift" on a music platform, **you're using similarity search without even knowing it.** At its core, similarity search helps find items that are alike.

### What is similarity search and why is it important?

Think of similarity search as a smart librarian. This librarian doesn't just look at book titles, but understands what each book is about. When you ask for "books like Harry Potter," this librarian knows to recommend other fantasy novels.

These novels would have coming-of-age stories and magical schools. **This is exactly what similarity search does – it understands the essence of what you're looking for and finds items that match that essence.**

### How similarity search works with vector embeddings

To understand how computers can find similar items, imagine turning everything into a list of numbers. When you take a photo, write a sentence, or record a song, similarity search converts it into a special list of numbers. This list is called a [vector embedding](https://meilisearch.com/blog/what-are-vector-embeddings).

**These numbers capture the important features of the item.** For a photo, it might include information about colors, shapes, and objects present.

Think of these vectors like coordinates on a map. You can find nearby cities by looking at their location on a map. **Similarity search finds similar items by looking at how close their vectors are to each other.** Items that are similar will have vectors that are close together in this mathematical space.

### Key differences between similarity search and nearest neighbor search

While these terms are often used interchangeably, they serve slightly different purposes. Nearest neighbor search is like using a measuring tape to find the closest points to where you're standing. **It's precise but can be slow if you have to measure the distance to every single point.**

Similarity search, on the other hand, is more like asking for directions. It might not give you the absolute closest match, but it's faster and usually good enough.

### What is the role of similarity search in AI-driven applications?

AI applications rely heavily on similarity search to make sense of vast amounts of information. When you use a chatbot that answers your questions, it's likely using similarity search to find relevant information in its knowledge base.

<Callout intent="note">
	
Curious about how to implement similarity search in a real application? Learn [how to build a RAG system with similarity search](https://www.meilisearch.com/blog/how-to-build-rag) to enhance your AI applications.
	
</Callout>

## Real-world applications of similarity search

Similarity search has changed how we interact with digital content. It powers many features we use daily. Let's explore some practical applications that showcase its impact across different domains.

### Content-based retrieval in multimedia systems

When you upload a photo to Google Images and ask "find similar images", you're experiencing content-based retrieval.

Pinterest, for instance, uses this technology to help users discover visually similar pins. If you find a cozy living room design you like, the platform can instantly show you dozens of similar interior designs by comparing their visual embeddings.

![Green kitchen cabinets interior design](https://unable-actionable-car.media.strapiapp.com/Green_kitchen_cabinets_interior_design_980c857e9b.avif)

These systems break down images and videos into vector embeddings that capture visual elements like colors, shapes, and patterns.

### Making recommendations more personal and accurate

Recommendation systems have come a long way from simple "users who bought X also bought Y" suggestions. **Modern platforms use similarity search to create rich, personalized experiences.**

Take Spotify's Discover Weekly playlist. It combines your listening history, favorite genres, and even the acoustic properties of songs you love to recommend new music you're likely to enjoy.

![Spotify Discover Weekly ad](https://unable-actionable-car.media.strapiapp.com/Spotify_Discover_Weekly_ad_46d6a0219c.avif)

E-commerce sites like Amazon use multi-modal similarity, combining different types of data to improve recommendations. They might consider:

- Product descriptions and reviews (text data)
- Product images (visual data)
- Purchase patterns (behavioral data)
- Price ranges and categories (numerical and categorical data)

By analyzing all these dimensions together, they can suggest products that truly match what you're looking for, rather than just showing popular items in the same category. **This creates a better experience for the user.**

### Real success stories from the field

Similarity search in healthcare, beyond imaging, aids diagnosis. Hospital systems use it to match patients with similar symptoms and histories by vectorizing patient data.

[Research shows its utility in predictive analytics](https://pmc.ncbi.nlm.nih.gov/articles/PMC5357318/), improving accuracy in areas like diabetes prediction. This supports personalized treatment and outcome prediction for more effective care.

### Powering modern AI applications

The rise of LLM and retrieval-augmented generation (RAG) systems has created new applications for similarity search. When you chat with an AI assistant that needs to pull relevant information from a knowledge base, similarity search is working behind the scenes to find the most relevant content to inform its responses. **This is a game changer for AI applications.**

![RAG workflow diagram explanation](https://unable-actionable-car.media.strapiapp.com/RAG_workflow_diagram_explanation_53f23eb06c.avif)

For example, when a customer service chatbot needs to answer a specific question about a product, it uses similarity search to find the most relevant product documentation, support tickets, and FAQ entries. This helps the bot provide accurate, contextual responses rather than generic answers.

## Distance metrics: the heart of similarity search

When searching for similar items, we need a way to measure how close or far apart vectors are from each other. This is like using a ruler to measure physical distances.

### Understanding the main distance metrics

The choice of distance metric can make or break your similarity search. Think of it like choosing the right tool for the job. You wouldn't use a hammer to cut wood, right? **The three most popular distance metrics each have their sweet spots.**

![Euclidean distance diagram](https://unable-actionable-car.media.strapiapp.com/Euclidean_distance_diagram_e711c166fc.avif)

Euclidean distance works like a straight line between two points. It's the distance "as the crow flies". This metric shines when working with physical measurements or when the magnitude of your vectors matters. For example, if you're building a system to find similar house prices, Euclidean distance would be a great choice because the actual numerical differences matter.

![Cosines distance illustration](https://unable-actionable-car.media.strapiapp.com/Cosines_distance_illustration_f59696183b.avif)

Cosine similarity, on the other hand, cares about the angle between vectors, not their length. It's perfect for text search because it can tell if two documents are about the same topic even if one is much longer than the other. It helps deliver more relevant results regardless of text length.

![Manhattan distance L1 distance](https://unable-actionable-car.media.strapiapp.com/Manhattan_distance_L1_distance_828cda7a2a.avif)

Manhattan distance (also called L1 distance) measures distance as if you're walking through city blocks. You can only move horizontally or vertically. It's particularly useful when dealing with grid-like data or when you want to give equal weight to all differences between vectors.

### Choosing the right metric for your data

| Metric | Use Case | Key Benefit |
| --- | --- | --- |
| Cosine Similarity | Text Search | Handles varying document lengths; focuses on meaning. |
| Euclidean Distance | Image Search, Sensor Data | Measures numerical differences; scale is meaningful. |
| Manhattan Distance | Categorical Data (User Preferences, etc.) | Treats each dimension independently; robust to outliers in features. |

Here's a practical way to think about it.

For text search, cosine similarity is usually your best bet. It handles documents of different lengths well and focuses on the meaning rather than the size. This is especially important when searching through articles, product descriptions, or user queries.

For image search, Euclidean distance often works better because the actual numerical differences between pixel values or image features matter. The same goes for sensor data or any numerical measurements where the scale is meaningful.

For categorical data (like user preferences or product attributes), Manhattan distance can be more appropriate. It treats each dimension independently and doesn't get thrown off by large differences in individual features.

### Handling mixed data types

Real-world applications often deal with multiple types of data at once. For instance, an e-commerce search might need to consider both product descriptions (text) and product images. **In these cases, you can use hybrid approaches.**

1. Calculate similarities separately using appropriate metrics for each data type
2. Combine the results using weighted averages
3. Normalize the scores to ensure fair comparison

This flexible approach allows you to fine-tune the importance of different features. You might want product images to carry more weight than text descriptions when searching for clothing items.

### Impact on search performance

The choice of distance metric doesn't just affect accuracy. **It can significantly impact search speed too.** Euclidean and Manhattan distances are generally faster to compute than cosine similarity, but modern search engines like Meilisearch optimize these calculations so you rarely need to worry about performance differences.

What matters more is choosing a metric that matches your data and use case. A faster metric that gives less relevant results isn't a good trade-off. Focus first on what makes sense for your users and their search needs, then optimize for performance if necessary.

## Algorithms that power similarity search

Now we understand how distance metrics help measure similarity between vectors. Let's then explore the algorithms that make searching through these vectors efficient and scalable. They are specifically designed to handle vector-based searches while balancing speed and accuracy requirements.

Different search algorithms handle distance calculations in various ways - some compare against every vector for perfect accuracy, while others use clever shortcuts to speed up the process. The choice of algorithm often depends on your dataset size, the dimensionality of your vectors, and whether you need exact or approximate results.

Let's examine the main approaches to similarity search, from basic exact matching to sophisticated approximate methods.

### Exact vs. approximate: finding the best match

![](https://unable-actionable-car.media.strapiapp.com/knn_algorithm_illustration_4f027a5683.avif)

The k-Nearest Neighbors (k-NN) algorithm finds the *exact* closest matches. It compares your search to every single item. While accurate, this is slow with large datasets. **k-NN isn't practical when speed is important.** Imagine comparing one book to millions!

![ANN algorithm illustration](https://unable-actionable-car.media.strapiapp.com/ANN_algorithm_illustration_ad3c4df68e.avif)

Approximate Nearest Neighbors (ANN) algorithms are a faster alternative. They make educated guesses to find good matches quickly. **ANN might miss the *absolute* best match sometimes, but it's much faster and accurate enough for most uses.**

### Space partitioning and graph navigation

Space partitioning methods, like KD-trees and Voronoi diagrams, organize data for faster searching. KD-trees divide the search space into smaller regions. Voronoi diagrams divide space based on proximity to certain points. **These are good for simpler data, but less useful with complex, high-dimensional data.**

[Hierarchical Navigable Small World](https://en.wikipedia.org/wiki/Hierarchical_navigable_small_world) (HNSW) is a cutting-edge algorithm for similarity search. It creates a network of connections between data points. **HNSW is great for complex data used in modern AI.** It can search millions of items quickly while providing relevant results, ideal for things like [semantic search](https://meilisearch.com/blog/semantic-search) and [recommendation systems](https://meilisearch.com/blog/personalization-recommendations).

## The future of search: embracing semantic intelligence

Similarity search is transforming information retrieval by connecting traditional methods with advanced semantic understanding.

**This technology uses vector embeddings and sophisticated algorithms for more intelligent, context-aware search experiences.** As AI evolves, similarity search will be essential in making search more intuitive, precise, and meaningful across many domains and applications.

Similarity search: a guide to vector-based retrieval  

# Meilisearch AI launch week recap

As the first quarter of 2025 closes, we've completed our first [Meilisearch launch week](https://www.meilisearch.com/launch-week), marking our most significant product evolution since introducing Meilisearch Cloud. This update brings comprehensive AI-powered search capabilities to all users.

Traditional keyword search has limits. It matches characters, not intent—often returning results that miss the mark. [AI-powered search changes that](https://www.meilisearch.com/blog/introducing-meilisearch-ai). By understanding meaning, not just words, it delivers smarter, more intuitive results.

## What we announced: Day by day

#### Day 1: Semantic and hybrid search now GA

We kicked off the week by announcing that our AI capabilities, including semantic search and hybrid search, are now generally available:

- **Semantic search**: AI-powered understanding delivers more relevant results beyond simple keyword matching
- **Hybrid search**: Combines traditional full-text reliability with semantic intelligence
- **Performance optimization**: All this intelligence with sub-50ms response time

We launched semantic and hybrid search as generally available on the Meilisearch platform. Now your search results understand context, personalize experiences, and power better recommendations, whether you're building an e-commerce site, a knowledge base, a media and news archive, or a SaaS app.

These capabilities significantly reduce failed searches and "no results" responses. [E-commerce platforms report higher conversion rates](https://www.meilisearch.com/blog/Agora) as shoppers find products using natural language, while [app builders](https://www.meilisearch.com/blog) see increased engagement as users discover precisely what they need, even with imprecise queries.

**📹 [Watch our Head of Product, Guillaume, explain the release](https://www.youtube.com/watch?v=3JaDBEQNLZc&t).**


#### Day 2: Multiple embedder options and vector storage

On day two, we made vector search more flexible with multiple embedder choices—eliminating the need for separate vector database infrastructure:

- **Choose your embedder**: OpenAI, Hugging Face, Ollama, or bring your own
- **documentTemplate feature**: Better prompting for improved results
- **Binary quantization**: Optimize large databases for speed while maintaining result quality

Our flexible embedder options address critical infrastructure and cost concerns. Organizations can choose solutions that balance budget constraints, performance requirements, and proprietary needs. This approach eliminates the expense of maintaining separate vector databases and reduces operational complexity.

**📹 [Watch the walkthrough – Our Sr. Product Manager, María, explains embedders and best practices.](https://youtu.be/0tPsBrmtf3o?si=6KAyuREv0DYVfCZD).**


#### Day 3: Multi-modal search capabilities

The middle of the week brought multi-modal search, allowing searches beyond just text. This opens new possibilities for industries with diverse content types:

- **Flexible input types**: Search across text, images, and more
- **AI-powered semantics**: Smarter results across content types
- **Real-world implementations**: learn more about how early adopters like [Agora](https://www.meilisearch.com/blog/Agora) and [MotionElements](https://www.meilisearch.com/blog/motionelements) already using these capabilities
    
    ![MotionElements_multimodal.gif](https://unable-actionable-car.media.strapiapp.com/Motion_Elements_multimodal_6cfa386d8d.gif)

#### Day 4: Personalized recommendations

Thursday's focus was on helping users discover relevant content faster:

- **/SIMILAR documents**: Find content similar to what you're viewing
- **Engagement improvements**: Smart, similarity-based product and document recommendations
- **Seamless user experience**: Help users find what they need (and didn't know about)
    
    ![day 4_recommendations.gif](https://unable-actionable-car.media.strapiapp.com/day_4_recommendations_e67bc9c528.gif)

#### Day 5: Wrapping up and looking ahead

https://youtu.be/I42NbytO0hQ

We closed the week with a recap and vision for the future with our CEO, Quentin.

At Meilisearch, our mission is to make the best search experience accessible to everyone. We want to deliver the best technologies and combine them with an amazing user experience. This week, we’ve taken a big step forward and are closer to this vision, helping search results understand context, offer personalized experiences, and enable smarter recommendations.

In the coming months and quarters, you'll see exciting new features—from better search analytics and more insights for taking action on your business, to more customization capabilities for tailored experiences, but also we’ll continue advancing on exclusive capabilities like [built-in RAG](https://www.meilisearch.com/blog/how-to-build-rag?ref=dailydev) or [image search](https://www.meilisearch.com/docs/learn/ai_powered_search/image_search_with_user_provided_embeddings) to make them work out of the box, with little to no configuration required.

**📹 [Watch Meilisearch AI launch week wrap up with CEO](https://youtu.be/0tPsBrmtf3o?si=6KAyuREv0DYVfCZD).**

**📖 Read Quentin’s [blog](https://www.meilisearch.com/blog/introducing-meilisearch-ai) on building the future of search with Meilisearch AI.**
## Join us on this journey

As we celebrate reaching **50,000 GitHub stars,** we're just getting started on our mission to redefine search. Whether you're new to Meilisearch or already part of our community, these new capabilities unlock more intuitive, personalized search experiences.

To mark the occasion, we’re giving new subscribers an exclusive gift: use the code **`MEILIAI`** and **get [2 months of Meilisearch](https://help.meilisearch.com/en/article/meilisearchs-promo-discount-code-program-1ykitx7/), including AI-powered searches, completely free**. It’s the perfect time to start building the search experience you’ve always wanted.

Meilisearch AI launch week recap

[Agora](https://www.searchagora.com/) chose Meilisearch to power lightning-fast, intuitive, and scalable search experiences across millions of ecommerce products.

Agora is a revolutionary ecommerce search platform aiming to become the decentralized "world store,” which aggregates products from thousands of smaller online shops into one cohesive marketplace.

It's built for shoppers looking to effortlessly discover, compare, and purchase products from multiple vendors within one unified cart. Agora has ambitious scaling plans, aiming to index up to 1 billion products by the end of 2025. To do this, the team needed a blazing-fast solution capable of delivering relevant search results instantly and at scale.

>“Meilisearch's combination of semantic search and speed was the best we’ve seen on the market” - Param Jaggi, CEO & Founder at Agora

![Agora gif.gif](https://unable-actionable-car.media.strapiapp.com/Agora_gif_eca98f7605.gif)

**Challenge: building lightning-fast, relevant search across diverse ecommerce platforms at massive scale**
-------------------------------------------------------------------------------

As Agora's dataset rapidly grew from millions toward a projected billion product listings, their initial search setup faced significant hurdles that threatened user experience and growth:

- **Challenge: aggregating and standardizing diverse data sources** 
Agora pulls product data from thousands of Shopify and WooCommerce stores, each with unique data structures. To create a unified shopping experience, Agora faced a challenge of standardizing complex data into one comprehensive schema that could be seamlessly indexed and searched.
- **Challenge: providing immediate, relevant search results.**
Initially, Agora used Mongo Atlas Search, where search queries often took 5 to 10 seconds. It significantly deteriorated user experience and conversions. Agora required a solution delivering instant results to ensure users discovered the right products rapidly and confidently, even for complex or ambiguous queries.

- **Challenge: maintaining control and flexibility over AI embeddings.** 
Previous search solutions did not provide Agora with sufficient control over their own vector embeddings, limiting flexibility and customization. Agora sought a solution enabling them to leverage their embeddings, fine-tune relevancy, and experiment freely with semantic and keyword search combinations.

**Solution: Leveraging Meilisearch’s hybrid AI search engine for exceptional ecommerce discovery**
-------------------------------------------------------------------------------

Agora turned to Meilisearch for its powerful combination of speed, intuitive semantic search capabilities, and unmatched ease of customization. Meilisearch addressed Agora’s challenges head-on by:

- **Lightning-fast search-as-you-type results:** Meilisearch processes searches in under 100 milliseconds, allowing real-time, dynamic result updates as users type, significantly increasing relevancy and conversion rates.
- **Hybrid semantic & keyword search** powered by vector embeddings: Agora leverages Meilisearch’s AI-powered semantic search capabilities by integrating OpenAI’s CLIP model embeddings. This allowed customers to intuitively search for products using both textual input as well as search by image.
- **Customizable schema and indexing flexibility**: Meilisearch’s loosely typed document structure enabled Agora to standardize and index complex product data across diverse sources.

**Implementation: how Agora built a blazing-fast product search with Meilisearch**
-------------------------------------------------------------------------------

To deliver fast, relevant, and intuitive search experiences, Agora implemented Meilisearch as follows:
1. **Multi-source product crawling:** using its own custom crawler called “Hermes”, Agora scraped and standardized product data from diverse Shopify and WooCommerce storefronts into a single structured format optimized for indexing with Meilisearch.
1. **Leveraging advanced embeddings for semantic discovery:** Agora then integrated Cohere for semantic text embeddings and OpenAI’s CLIP model to generate powerful semantic embeddings for products' images. These embeddings augmented keyword searches, enabling a hybrid search that blended semantic depth and keyword precision, drastically enhancing result quality.
3. **Real-time search optimization through filtering, sorting, and ranking rules:** Meilisearch helped Agora refine search relevancy by balancing semantic and keyword search. This was essential for complex e-commerce queries and creative product names. By weighting semantic search, Agora prioritized relevant terms within context, ensuring accurate results even with unconventional naming.

**Results: transforming ecommerce search with AI-powered discovery**
-------------------------------------------------------------------------------
Adopting Meilisearch helped Agora’s team overcome their biggest search challenges, significantly improving their user experience and driving measurable performance gains.

Some of the results include:
- **Dramatically improved search speeds**: Search queries, previously averaging around 2 seconds, now return in approximately 300 milliseconds (with core Meilisearch responses at just under 100 milliseconds). Agora's team is confident in further reducing this down below 200 milliseconds, significantly improving user satisfaction and search conversion rates.
- **Increased conversions and customer trust:** Since implementing Meilisearch, Agora saw a 1.7x to 2x increase in weekly user retention and a notable uptick in order conversions, directly attributed to the improved search experience and consistently relevant results.
- **Enhanced discovery with semantic embeddings:** Agora’s users can intuitively search through both text and images, easily discovering desired products using natural language or visual descriptions. Embeddings also unlocked new capabilities like strong similarity-based recommendations, expanding product discovery avenues for shoppers.

**Conclusion: empowering Agora’s ambitious vision for a unified "World Store“**
-------------------------------------------------------------------------------

Agora's mission to create a unified marketplace experience across millions—soon billions—of products demands an exceptional search solution that balances speed, relevance, and flexibility.

Meilisearch proved to be the perfect choice, enabling Agora to seamlessly scale its product catalog while maintaining outstanding search performance, intuitive semantic discovery, and ultimate flexibility.

With Meilisearch’s powerful hybrid search engine, Agora is confidently positioned to deliver a truly transformative e-commerce experience.


Scaling to billions: How Agora elevates e-commerce search with Meilisearch AI

When [Bookshop.org](http://Bookshop.org)'s small engineering team struggled to push past a 14% search conversion rate, they faced a familiar dilemma. Despite having Elasticsearch—a powerful but complex search solution—their eight-person team was spending more time maintaining search infrastructure than improving their core platform. They needed something different: a solution that could deliver better results with less overhead.

Their challenge reflects a broader reality in modern software development. Every click, every search query, and every page load directly impact business metrics, yet the tools meant to optimize these experiences often become bottlenecks themselves. For engineering teams, especially smaller ones, the complexity of search infrastructure can become a _hidden tax_ on innovation and growth.

This is where Meilisearch comes in. Built to work seamlessly out of the box, it's the solution that lets developers focus on building rather than maintaining. Today, we're taking that mission further with [the launch of Meilisearch AI](https://www.meilisearch.com/launch-week)—our next step in making it radically easier for developers to build world-class search experiences.

In this article, we'll explore how modern search has evolved beyond simple queries, examine the technical challenges facing development teams, and introduce Meilisearch's innovative solutions—including our new AI capabilities. You'll learn how our approach to developer experience is transforming search implementation, and discover our roadmap for making advanced search features accessible to teams of all sizes.

## Redefining modern search

Modern search experiences extend far beyond the traditional search bar, encompassing two key facets: search and discovery. 

Search can take the appearance of inputs, menus, filters, or the trending UX of conversational search (i.e. chatbots). Discovery, on the other hand, plays a key role in navigating content and suggesting recommendations.

Across industries, we see search and discovery most prevalent as follows:

- **Traditional full-text search & semantic search:** standard in all industries ranging from SaaS to ecommerce, marketplaces, and content platforms
- **Conversational search:** prominent in RAG (Retrieval-Augmented Generation) and AI-enhanced search interfaces like Perplexity
- **Discovery for navigation:** common in social media and content platforms where a feed is the primary navigation
- **Recommendations:** tie in more critically with business value in ecommerce and content platforms

For early-stage startups and SMBs, implementing these search capabilities presents significant technical challenges. The complexity multiplies when adding AI integration and personalization layers. Addressing these pain points diverts the workforce from building core product value. Meilisearch is here to relieve you from this burden, allowing you to focus on what really matters. 

We want Meilisearch built-ins to handle modern use cases like [document embedding](https://www.meilisearch.com/docs/learn/ai_powered_search/document_template_best_practices). Complex setups—such as managing [heavyweight solutions like Elasticsearch](https://www.meilisearch.com/blog/meilisearch-vs-elasticsearch), [tweaking a Postgres instance](https://www.meilisearch.com/blog/postgres-full-text-search-limitations), or handling hybrid search—are additional integration burdens. This ignores the security consideration related to exposing front-end accessible APIs or the analytics necessary to collect business KPIs—things that don’t come packaged with custom-made solutions.

![bookshop_meilisearch.gif](https://unable-actionable-car.media.strapiapp.com/bookshop_meilisearch_9498082583.gif)

These DX considerations impact business metrics down the road as engineering bottlenecks lead to longer time-to-market, hindering growth for companies that want to move fast. In ecommerce, for example, search accuracy directly translates to conversion rate, as shown by how we helped [Bookshop increase search-based purchases by 43%](https://www.meilisearch.com/blog/bookshop-increases-search-based-purchases). On SaaS and content platforms, discovery enables engagement and adoption growth, both of which are key to keeping churn in check. 

## Leveraging DX to move mountains

Meilisearch is a no-configuration, plug-and-play solution that you can install and forget. At the heart of any search tool—and often the biggest challenge for developers—is ensuring relevant results. **CascadeRank** is our answer to [relevancy fine-tuning](https://www.meilisearch.com/docs/learn/relevancy/relevancy). Our algorithm uses human-readable ranking rules that enable configuration and easy customization. With Meilisearch, simplicity is key, allowing you to set up and focus on your core tasks. 

Though customization is easy, Meilisearch comes with strong defaults to handle common requirements out of the box—things like being typo-tolerant and prioritizing word proximity. Indeed, we’ve seen [Minipouce get better relevancy with Meilisearch’s default configuration than their previous Typesense setup](https://www.meilisearch.com/blog/minipouce-top-grade-customer-experience), leading to a 5x search volume increase. 

Scalability is a concern for small teams and enterprises alike. Performance is the name of the game for user-facing search—both for search latency and indexing time. As demonstrated by [Podscan.fm](http://Podscan.fm) [searching across 600GB of data](https://x.com/arvidkahl/status/1807889668748275802), search remains fast against large datasets. Our [next-generation indexer boasts 4x faster updates](https://www.meilisearch.com/blog/introducing-indexer-2024), making Meilisearch suitable at scale for write-intensive workflows, too.

[![arvid-kahl-tweet.png](https://unable-actionable-car.media.strapiapp.com/arvid_kahl_tweet_0c41da4824.png)
](https://x.com/arvidkahl/status/1807889668748275802)

When it comes to security, no compromise is possible. So Meilisearch provides fundamental security features to handle simple and complex use cases. Meilisearch facilitates [API key management](https://www.meilisearch.com/docs/learn/security/basic_security) to facilitate multi-environments. The built-in [multi-tenancy features](https://www.meilisearch.com/docs/learn/security/multitenancy_tenant_tokens) also enable more advanced security patterns, allowing developers to customize their access rules.

Using our expertise in search requirements across various customers, we built [Meilisearch Cloud](https://www.meilisearch.com/cloud). It’s the quickest way to get started. In addition to freeing teams from managing infrastructure, it integrates analytics and monitoring, making it [the preferred production platform for companies like HitPay](https://www.meilisearch.com/blog/hitpay-streamlines-search). Open-source Meilisearch can be self-hosted via Docker and a plethora of hosting platforms. 

This feature set positions Meilisearch well as a simple solution to get the best of modern search experience. This includes benefiting from an easy LLM integration to leverage AI-powered solutions.

## Future-proofing your search strategy

With native [hybrid and vector search](https://www.meilisearch.com/docs/learn/ai_powered_search/getting_started_with_ai_search) integration, Meilisearch makes it simple to implement AI-powered search capabilities. But we’re not stopping there—our 2025 roadmap is packed with features enabling businesses to uncover more value from AI while sustaining performance and scalability.

### Improved conversions with personalization

[Our research shows that personalization significantly improves conversion rates](https://www.meilisearch.com/blog/personalization-recommendations) across diverse business types. We aim to release a personalization framework enabling the collection of end-user context on each request. This comprehensive context management system allows developers to leverage user preferences, recent interactions, and behavioral patterns to further customize the relevancy of results on a per-user basis. 

### Retrieval augmented generation without overhead 

Although Meilisearch already enables [building retrieval-augmented generation (RAG)](https://www.meilisearch.com/blog/how-to-build-rag), we want to integrate built-in RAG capabilities in the engine, seamlessly combining full-text search, vector search, and LLM interaction. Then, we want to tackle inherent problems in state-of-the-art RAG applications by tackling the orchestration of filtering, ranking, and results merging in RAG scenarios. Of course, we want this to come with strong defaults, making it easy for developers to get started.

### Advanced ranking integrations

Our goal is to provide a native way for users to leverage AI reranking models to improve the relevancy of results. This provides a more flexible approach to configuring result relevancy by allowing real-time results reranking based on user context and behavior. We also want to enable organizations to integrate with their own custom ranking models to allow for further dynamic adjustment of results based on hand-picked factors.

### Reliability as you scale

We also plan to improve Meilisearch’s infrastructure to handle additional use cases better. In particular, we want to continue improving the distributed system capabilities for better scaling and high availability by building upon our sharding experiments with [federated remote search](https://www.meilisearch.com/docs/learn/multi_search/implement_sharding). Another layer of performance is integrating more sophisticated caching mechanisms to improve out-of-the-box performance. Monitoring and observability are also continuously improved based on user feedback.

## Meilisearch AI, refreshingly simple search DX

The release of Meilisearch AI marks a new age for Meilisearch users, fully integrating with the modern landscape of AI-powered applications. But it’s also a continuous fulfillment of our core value—making good search UX easy to build. We believe focusing on developer experience enables engineers to achieve faster time-to-value and maintain focus on their core business.

Meilisearch ships frequently to keep packaging built-in features, enabling fast-shipping companies to do what they do best. For more information on the Meilisearch AI release, consult our [launch week page](https://www.meilisearch.com/launch-week). 

Start with a [14-day free trial of Meilisearch Cloud](https://cloud.meilisearch.com/) or [request a demo](https://dub.sh/meili-sales) to discuss your needs with our search experts.

Building the future of search with Meilisearch AI

**[Rayon](https://rayon.so/) is a cloud-based design platform for architecture and interior design that enables teams to collaborate in real-time.**

With millions of monthly searches and users in over 100 countries, Rayon needed a fast, multilingual search solution to keep up with explosive data growth. They turned to Meilisearch to power their semantic search and scale seamlessly.

>“Meilisearch is very powerful for our needs, especially its semantic search, which effortlessly helps users find relevant objects - even without precise queries.” - said Bastien Dolla, CEO & Co-founder at Rayon.

**Challenge: delivering fast, intuitive, and scalable search experience, while accommodating rapid growth.**
-------------------------------------------------------------------------------

With a rapidly growing user base in over 100 countries, Rayon needed to address critical challenges, including improving search functionality and supporting the rapid expansion of its database and usage while enhancing the overall user experience.

**Challenge: Multilingual search for a global audience.**
Traditional keyword-based search systems usually require users to know the exact name of an asset, often in English. Rayon’s non-native English-speaking users make up a significant portion of its global audience, so the platform needed to support searches in multiple languages in order to eliminate friction.

**Challenge: Rapid data expansion.**
Rayon’s platform offers access to over 4,000 assets, which represent real-world objects like furniture or textures. Users can also create and publish their own libraries, which leads to the rapid growth of the repository. With approximately one terabyte of new data added weekly, conventional tools simply couldn’t keep pace.

**Challenge: Scaling demand.**
The platform handles 2 million searches monthly, and expects to hit 20 million by year’s end. This necessitated a high-performance, scalable solution.

**Solution: creating an AI-powered search experience for Rayon’s global user base.**
-------------------------------------------------------------------------------

Rayon turned to Meilisearch to address its challenges with multilingual search, scaling, and efficient data management. These were addressed by:

- Leveraging AI-powered [semantic search](https://www.meilisearch.com/blog/semantic-search) for intuitive discovery: rather than relying on exact keywords, Meilisearch understands intent. For example, whether a user searches for “Scandinavian style” or “modern sofa with clean lines,” relevant results are quickly delivered.
- Enabling multilingual search: In addition to understanding the natural meaning of a query, Meilisearch supports non-English language searches, leveraging OpenAI embeddings to eliminate language barriers. For example, a search for 'chaise' in French or 'silla' in Spanish will return relevant chair designs, regardless of the language used.
- Scalable architecture for rapid growth: By indexing only design asset libraries and using token-based access control, Meilisearch efficiently manages data growth while ensuring stable performance and secure, authorized access.
- Open-source flexibility and Rust compatibility: The open-source origins of Meilisearch ensures that Rayon retains control over its search infrastructure, which provides them the flexibility to deploy their own instance if needed. Compatibility with Rust, the programming language Rayon is built on, is an important commonality to Rayon as they have a strong Rust team, making it easier for them to integrate and work with Meilisearch

**Implementation**
-------------------------------------------------------------------------------

Rayon implemented an efficient indexing process to ensure users could quickly and accurately find the design assets they needed:
1. Library processing and SVG generation: When users publish a library of design assets, Rayon processes the files and generates SVG images for each element. These images serve as visual representations of the assets, making it easier for users to identify what they are searching for.
2. Embedding generation with OpenAI: Rayon uses OpenAI’s advanced AI models to generate embeddings for each design asset. These embeddings capture the semantic meaning of the asset, enabling Meilisearch to perform highly accurate searches based on natural language descriptions.
3. Token-based access control: Meilisearch’s token system allows Rayon to maintain strict control over data visibility while still providing a seamless search experience.

**Results**
-------------------------------------------------------------------------------
By solving key challenges, Meilisearch has become essential to Rayon’s mission, delivering faster searches, happier users, and a platform built for the future:
- **Multilingual search for a global audience.** Designers can now search for assets in their native language, eliminating the need for English-only queries.
- **Faster, more intuitive search:** Meilisearch’s high-performance engine enables users to find design assets in seconds using natural language, without needing exact names.
- **AI-powered discovery** – With [OpenAI embeddings](https://www.meilisearch.com/docs/learn/ai_powered_search/getting_started_with_ai_search#generate-embeddings-with-openai), Rayon's search goes beyond keywords, allowing users to find assets based on descriptions, even without precise terminology.
**Scalability for rapid growth** – Meilisearch indexes 250,000–350,000 searchable items, supporting both Rayon’s library and user-generated content. With 1TB of new data added weekly, it seamlessly handles growing search volumes—scaling from 2M to an expected 20M searches per month by year-end.

**Conclusion: Solidifying Rayon’s position as a leader in cloud-based design tools.**
-------------------------------------------------------------------------------
Meilisearch has become essential to Rayon's mission by solving multilingual access, scalability, and asset discovery challenges.
As Rayon continues to innovate, the team plans to explore new capabilities, such as visual search and deeper data and [RAG](https://www.meilisearch.com/blog/how-to-build-rag) integration, to evolve into a comprehensive platform for architects and designers.
Looking ahead, their vision goes beyond visual assets and drawings, aiming to integrate procurement details, material specifications, and other essential data into a **truly comprehensive design hub.**




How Rayon leverages AI semantic features for a seamless design UX 

What are vector embeddings? A complete guide [2025]

Semantic search transforms how we find information online by focusing on understanding the meaning behind the words and phrases in user searches rather than relying solely on word-for-word matching.

Using machine learning (ML) and natural language processing (NLP), it decodes intent, context, and relationships between words, thus allowing for more conversational queries. This makes it a vital tool for businesses needing fast, accurate, and relevant results from large or complex data sets.

Embedding models are ML techniques that render words and phrases into complex numerical representations, which are then classified based on their context and relationships. However, not all models are the same. Accuracy, speed, indexing efficiency, and pricing can vary significantly between them.

Choosing the right semantic search model is crucial for delivering precise, efficient, and scalable search experiences for specific tasks.

Understanding the trade-offs—whether for multilingual support, search latency, or computational efficiency—will help you implement a high-performing semantic search system for your brand.

## What is semantic search?

Semantic search is an advanced AI-powered approach to information retrieval that focuses on NLP, ML, and knowledge representation to deduce the intent and contextual meaning of user queries.

Traditional keyword-based search engines match exact strings of words, but that’s not always enough. Semantic search goes further by incorporating **named entity recognition**, **relationships between words**, and **contextual disambiguation** to deliver more relevant results.

Semantic search engines can understand synonyms, paraphrased queries, and even infer implicit meanings exclusively because they use deep learning models such as transformers.

Given its ability to enhance the search criteria, [AI-powered search](https://www.meilisearch.com/blog/ai-powered-search) is crucial for applications requiring faster, broader, and even less structured [data retrieval](https://www.meilisearch.com/blog/how-to-cache-semantic-search).

## Why is semantic search important?
Semantic search is important because it improves search capability and relevance, thus helping users find information faster and with fewer irrelevant results. It also reduces uncertainty by interpreting natural language queries more clearly.

Like similarity search, it allows users to phrase their questions conversationally and intuitively rather than using specific—sometimes even long-tail—keywords. Think of it like this: if you can ask a question from your neighbor on a bus, then you can ask it with a quick semantic search.

Additionally, businesses benefit from semantic search by **enhancing customer experiences**, **improving product discovery**, and **optimizing knowledge management systems**.

For example, e-commerce stores like Bookshop.org can [boost search-based sales by up to 43%](https://www.meilisearch.com/blog/bookshop-increases-search-based-purchases) by implementing semantic search tools like Meilisearch.

A core component of semantic search is embedding models through which similarity computations are made possible.

## What are embedding models?

Embedding models are machine learning models that transform words, phrases, or documents into dense numerical representations called embeddings.

These vector representations encode word relationships, which allow search engines to compare semantic meaning and context rather than rely solely on exact similarity between words.

For instance, embeddings generated for words like “**phone**,” “**mobile**,” and “**cell**” will be more closely aligned in vector space than, say, “**cell**” and “**bell**,” even though the latter two are closer in letter-by-letter similarity.

This allows a search engine to retrieve relevant results even if the exact keyword is absent, underpinning various NLP and LLM (large language model) applications.

### What’s the difference between semantic and search embeddings?

Semantic and search embeddings serve distinct purposes in NLP applications, particularly in semantic search.

**Semantic embeddings** capture the meaning and relationships between words, which is helpful for tasks like document classification, recommendation systems, language translation, and sentiment analysis. They do this by reflecting the words’ conceptual similarities, even if they are not used in the same context.

On the other hand, **search embeddings** are specifically optimized for retrieval tasks, thus ensuring queries and indexed documents align effectively in vector space to maximize search relevance.

Unlike generic semantic embeddings, search embeddings often incorporate domain- and background-specific optimizations to fine-tune the effectiveness of the specific retrieval activity.

For example, a semantic embedding model might learn that "mac" is related to pasta, the makeup company, and the Apple laptop computer. However, a _search embedding model_ trained on computers and hardware would prioritize the third meaning when processing search queries in that domain.

Embedding models serve as the backbone of semantic search by enabling similarity-based retrieval mechanisms that align with user intent.

## What is the role of embedding models in semantic search?

Embedding models power semantic search by converting text into a structured format that can be efficiently indexed and retrieved.

Instead of relying on exact keyword matches, these models compare **query** embeddings with **indexed** document embeddings. This enables nearest-neighbor searches and significantly improves retrieval accuracy over traditional algorithms that depend on keyword matching.

These models often use transformer-based architectures such as BERT (Bidirectional Encoder Representations from Transformers), GPT (Generative Pre-trained Transformer), and their derivative neural networks to capture context-aware representations.

That’s how they can handle nuanced queries at scale. For instance, even though “cat” and “bat” sound (and look) similar, only one of them may be used in context as a conventional house pet.

Several fine-tuned and pre-trained models continue to emerge as industry standards for semantic search. Each improves precision, relevance, efficiency, and/or scalability in its own way.

## What are the most commonly used models for semantic search?

Different embedding models vary in terms of vector dimensionality, context length, and certain other performance characteristics.

In the context of LLMs, dimensionality points to the number of components in a vector, with each corresponding to an attribute of the encoded variable. Additionally, context length relates to the amount of text—usually measured in tokens—a model can “remember” and refer to at any given time.

To assess these variations, a series of benchmark tests were conducted using [Meilisearch](https://www.meilisearch.com), evaluating each model's effectiveness in real-world search scenarios that you will encounter.

These tests measured factors such as **retrieval accuracy**, **indexing speed**, and **query latency**, which is how they evaluated each model’s performance under specific search conditions.

| **Model/Service** | **Dimensions** | **Context Length** |
| --- | --- | --- |
| [Cohere embed-english-v3.0](https://docs.cohere.com/docs/embeddings) | 1024 | 512 |
| [Cohere embed-english-light-v3.0](https://docs.cohere.com/docs/embeddings) | 384 | 512 |
| [Cohere embed-multilingual-v3.0](https://docs.cohere.com/docs/embeddings) | 1024 | 512 |
| [Cohere embed-multilingual-light-v3.0](https://docs.cohere.com/docs/embeddings) | 384 | 512 |
| [OpenAI text-embedding-3-small](https://platform.openai.com/docs/guides/embeddings) | 1536 | 8192 |
| [OpenAI text-embedding-3-large](https://platform.openai.com/docs/guides/embeddings) | 3072 | 8192 |
| [Mistral](https://docs.mistral.ai/capabilities/embeddings/) | 1024 | 8192 |
| [VoyageAI voyage-2](https://www.voyageai.com/) | 1024 | 4000 |
| [VoyageAI voyage-large-2](https://www.voyageai.com/) | 1536 | 16000 |
| [VoyageAI voyage-multilingual-2](https://www.voyageai.com/) | 1024 | 32000 |
| [Jina Colbert v2](https://jina.ai/news/jina-colbert-v2-multilingual-late-interaction-retriever-for-embedding-and-reranking/) | 128, 96, or 64 | 8192 |
| [OSS all-MiniLM-L6-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2) | 384 | 512 |
| [OSS bge-small-en-v1.5](https://huggingface.co/BAAI/bge-small-en-v1.5) | 1024 | 512 |
| [OSS bge-large-en-v1.5](https://huggingface.co/BAAI/bge-large-en-v1.5) | 1536 | 512 |

Your choice of a model for semantic search will depend on a variety of factors, including, but not limited to, its **accuracy**, **computational efficiency**, and **cost**.

## What factors should be considered when choosing the best model for semantic search?

### 1. Results relevancy
Relevance is the cornerstone of effective semantic search, especially in achieving optimal user experience. The right model should balance precision, recall, and speed to ensure users receive highly relevant results without excessive noise.

This balance becomes especially important when comparing hybrid approaches like vector and full-text search. When selecting an embedding model, think about the following:

- multilingual support;
- handling multi-modal data;
- domain-specific performance.

In this context, bigger might not always mean better. While larger models often provide better accuracy, smaller models can offer competitive results at lower computational costs.

Moreover, effectively structuring data, such as with optimized document templates in Meilisearch, can improve search quality.

### 2. Search performance
Search latency is a critical factor in user experience. Search-as-you-type has become the standard for customer-facing applications as fast, responsive search results improve user engagement and retention. 

Local embedding models are ideal to achieve lightning-fast performance, as they eliminate the need for round trips to external services and reduce latency. If you must rely on remote models, hosting your search service close to the embedding service can minimize delays and improve user experience.

The table below showcases latency benchmarks for various local embedding models and embedding APIs. All requests originate from a Meilisearch instance hosted in the AWS London data center.

| **Model/Service** | **Latency** |
| --- | --- | 
| [Cloudflare bge-small-en-v1.5](https://huggingface.co/BAAI/bge-small-en-v1.5) | ±800ms |
| [Cloudflare bge-large-en-v1.5](https://huggingface.co/BAAI/bge-large-en-v1.5) | ±500ms |
| [Cohere embed-english-v3.0](https://docs.cohere.com/docs/embeddings) | ±170ms |
| [Cohere embed-english-light-v3.0](https://docs.cohere.com/docs/embeddings) | ±160ms |
| [Local gte-small](https://huggingface.co/models?search=gte-small) | ±20ms |
| [Local all-MiniLM-L6-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2) | ±10ms |
| [Local bge-small-en-v1.5](https://huggingface.co/BAAI/bge-small-en-v1.5) | ±20ms |
| [Local bge-large-en-v1.5](https://huggingface.co/BAAI/bge-large-en-v1.5) | ±60ms |
| [Mistral](https://docs.mistral.ai/capabilities/embeddings/) | ±200ms |
| [Jina colbert](https://huggingface.co/jinaai/jina-colbert-v2) | ±400ms |
| [OpenAI text-embedding-3-small](https://platform.openai.com/docs/guides/embeddings) | ±460ms |
| [OpenAI text-embedding-3-large](https://platform.openai.com/docs/guides/embeddings) | ±750ms |
| [VoyageAI voyage-2](https://www.voyageai.com/) | ±350ms |
| [VoyageAI voyage-large-2](https://www.voyageai.com/) | ±400ms |

Benchmarks conducted in Meilisearch highlight the stark differences in latency across various models, with local models achieving response times as low as 10ms while some cloud-based services reach 800ms.

### 3. Indexing performance

Efficient indexing is another critical factor for the scalability of your search solution. As expected, the time required to process and store embeddings varies widely among models. Notable metrics influencing processing time are API rate limits, batch processing capabilities, and model dimensions.

Local models without GPUs may experience slower indexing due to limited processing power, while third-party services vary in speed based on their infrastructure and agreements.

As mentioned earlier, minimizing data travel time between your application and the model can reduce latency and optimize indexing. Evaluating these factors ensures your chosen model and service can effectively meet your needs.

The benchmark below compares the indexing performance of a collection of 10k e-commerce documents (with [automatic embedding generation](https://www.meilisearch.com/docs/learn/experimental/vector_search?utm_campaign=vector-search&utm_source=blog&utm_medium=models-comparison#generate-auto-embeddings)).

| **Model/Service** | **Indexation time** | 
| --- | --- | 
| [Cohere embed-english-v3.0](https://docs.cohere.com/docs/embeddings) | 43s |
| [Cohere embed-english-light-v3.0](https://docs.cohere.com/docs/embeddings) | 16s |
| [OpenAI text-embedding-3-small](https://platform.openai.com/docs/guides/embeddings) | 95s |
| [OpenAI text-embedding-3-large](https://platform.openai.com/docs/guides/embeddings) | 151s |
| [Cloudflare bge-small-en-v1.5](https://huggingface.co/BAAI/bge-small-en-v1.5) | 152s |
| [Cloudflare bge-large-en-v1.5](https://huggingface.co/BAAI/bge-large-en-v1.5) | 159s |
| [Jina Colbert V2](https://huggingface.co/jinaai/jina-colbert-v2) | 375s |
| [VoyageAI voyage-large-2](https://www.voyageai.com/) | 409s |
| [Mistral](https://docs.mistral.ai/capabilities/embeddings/) | 409s |
| [Local all-MiniLM-L6-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2) | 880s |
| [Local bge-small-en-v1.5](https://huggingface.co/BAAI/bge-small-en-v1.5)| 3379s |
| [Local bge-large-en-v1.5](https://huggingface.co/BAAI/bge-large-en-v1.5)| 9132s |

Meilisearch benchmarks indicate that indexing times range from under a minute for optimized cloud-based solutions to several hours for certain local models without GPU acceleration.

These should be important points to ponder when you weigh the frequency and volume of data updates in your applications. This is because they directly impact how quickly your system can process frequent or large data updates, which is crucial for maintaining the performance and responsiveness of your search solution.

### 4. Pricing
Embedding model costs vary depending on the provider and usage model. While local models are free to run, they require computational resources, potentially necessitating GPU investment.

On the other hand, cloud-based services charge per million tokens (or thousand neurons for Cloudflare), with costs ranging from $0.02 to $0.18 per million tokens.

| **Provider** | **Price** |
| --- | --- | 
| Cohere | $0.10 per million tokens for Embed 3 |
| OpenAI | $0.02 per million tokens for text-embedding-3-small |
| | $0.10 per million tokens for text-embedding-ada-002 |
| | $0.13 per million tokens for text-embedding-3-large | 
| Cloudflare | $0.011 per 1,000 Neurons |
| Jina | $0.18 per million tokens |
| Mistral | $0.10 per million tokens |
| VoyageAI |$0.02 per million tokens for voyage-3-lite |
| | $0.06 per million tokens for voyage-3|
| | $0.12 per million tokens for voyage-multimodal-3 |
| | $0.18 per million tokens for voyage-code-3| 
| | $0.18 per million tokens for voyage-3-large |
| Local model | Free |

As such, analyze cost-effectiveness against search demand and performance requirements. It's often a good idea to start with a well-known model that's easy to set up and has robust community support. You can migrate to a cloud provider like AWS for improved performance when necessary. 

Alternatively, you can choose an open-source model to self-host, giving you more flexibility. Just be aware that optimizing local models for high volume may require scaling your infrastructure.

### 5. Other optimization techniques

To maximize search performance, [hybrid search approaches](https://www.meilisearch.com/blog/full-text-search-vs-vector-search) combining full-text and vector search can yield the best results. To refine semantic search performance, consider the following optimizations:

- Experiment with model presets, as some models allow tuning for query vs. document embedding, which can improve relevancy.
- Assess specialized models, especially those employing retrieval-augmented generation, as domain-specific models may offer superior results for particular use cases.
- Explore models that provide reranking functions that further enhance search precision.
- Test higher-tier accounts as premium tiers may provide faster processing and reduced rate limiting.
- Optimize data transfer with quantization options to reduce API response sizes and improve efficiency.

Carefully evaluating these factors will help you select the best semantic search model to meet your needs.

## The bottom line
Now that we’ve taken this journey together, let’s round up what we’ve discovered.

| **Model/Service** | **Dimensions** | **Context Length** | **Latency** | **Indexation time** | **Pricing (per million tokens)** |
| --- | --- | --- | --- | --- | --- | 
| [Cohere embed-english-v3.0](https://docs.cohere.com/docs/embeddings) | 1024 | 512 | ±170ms | 43s | $0.10|
|[Cohere embed-english-light-v3.0](https://docs.cohere.com/docs/embeddings)|384|512|±160ms|16s|$0.10|
| [OpenAI text-embedding-3-small](https://platform.openai.com/docs/guides/embeddings) |1536| 8192| ±460ms| 95s|$0.02|
|[OpenAI text-embedding-3-large](https://platform.openai.com/docs/guides/embeddings)| 3072|8192|±750ms|151s|$0.13|
|[Mistral](https://docs.mistral.ai/capabilities/embeddings/)|1024|8192|±200ms|409s|$0.10|
|[VoyageAI voyage-2](https://www.voyageai.com/)|1024|4000|±350ms|330s|$0.10|
| [VoyageAI voyage-large-2](https://www.voyageai.com/)|1536|16000|±400ms|409s|$0.12|
|[Jina Colbert v2](https://jina.ai/news/jina-colbert-v2-multilingual-late-interaction-retriever-for-embedding-and-reranking/)|128, 96, or 64|8192|±400ms|375s|$0.18|
|[OSS all-MiniLM-L6-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2)|384|512|±10ms|880s|Free|
|[OSS bge-small-en-v1.5](https://huggingface.co/BAAI/bge-small-en-v1.5)|1024|512|±20ms|3379s|Free|
|[OSS bge-large-en-v1.5](https://huggingface.co/BAAI/bge-large-en-v1.5)|1536|512|±60ms|9132s|Free|

Choosing the best model for semantic search depends on what (and how) you aim to achieve with your specific use case, budget, and performance requirements. For most cases, cloud-based solutions such as those provided by Cohere or OpenAI may be optimal.

As the needs of your organization grow, it might be worth the cost and effort to upgrade to a local or self-hosted solution. Understanding your own needs is paramount to making a well-informed decision.

If you’re unsure which model is best for you or if you’re looking for a tailored solution, contact a search expert at Meilisearch.

## Frequently asked questions (FAQs)

### How does semantic search differ from keyword-based search?
Semantic search focuses on understanding meaning, while keyword-based search relies on exact word matches.
### What are the key features of a good semantic search model?
A good model should offer high accuracy, low latency, efficient indexing, and cost-effectiveness.
### How do transformer-based models improve semantic search?
Transformers process text in context, capturing contextual relationships between words to enhance search relevance while reducing downtime.
### How do vector databases enhance semantic search?
Vector databases work to efficiently store embeddings, enabling fast and scalable search operations.
### What are some open-source models for semantic search?
Popular open-source semantic search models include all-MiniLM-L6-v2 and Universal Sentence Encoder.




How to choose the best model for semantic search

What is a vector database? What you need to know [2025]

**[MotionElements](https://www.motionelements.com/) is an AI-powered one-stop marketplace for video creators to access and customize creative assets. Time is money for MotionElements’ users, and search is essential in helping them quickly find what they need.** 

With a global user base speaking multiple languages, MotionElements relies on powerful multilingual search capabilities to deliver fast, accurate, and relevant results across diverse creative content.

![MotionElements_search.gif](https://unable-actionable-car.media.strapiapp.com/Motion_Elements_search_50cb3914a5.gif)
>“MotionElements’ federated hybrid search empowers creators to seamlessly explore diverse creative assets across our one-stop marketplace—all in their own language. By combining keyword search with zero-shot video search, visual & music similarity search, we deliver an intuitive, AI-enhanced discovery experience that fuels creative workflows.” - **Mark Sun, CTO at MotionElements**


**Challenge: Scaling search**
-------------------------------------------------------------------------------

Previously, MotionElements relied on Solr, which became increasingly difficult to maintain. With millions of new content pieces added each month, Solr consumed excessive RAM and struggled with scalability. MotionElements' search needs have also evolved, requiring support for keyword search, visual and music similarity vectors, and image embeddings—capabilities that are difficult to implement efficiently with Solr.


>_As our content grows, we need a search solution that can handle both traditional and AI-enhanced discovery while remaining scalable and performant,_ said Mark Sun, CTO of MotionElements.

![MotionElements_multimodal.gif](https://unable-actionable-car.media.strapiapp.com/Motion_Elements_multimodal_6cfa386d8d.gif)

**Solution: Choosing a superior price‑to‑performance ratio, scalability, and multilingual federated**
-------------------------------------------------------------------------------

- **Price‑to‑performance ratio:** MotionElements explored Elasticsearch, Algolia, and Typesense, and ultimately chose Meilisearch because of its price‑to‑performance ratio when handling large catalogs with millions of products.

- **Feature fit:** Meilisearch’s federated search capabilities and multilingual support were also key factors in MotionElements’ decision.

- **Addressing migration concerns**,  Mark Sun said:

>“Initially, we had concerns about migrating its image similarity vectors, but Quentin, the CEO of Meilisearch, provided invaluable support through a Proof of Concept (PoC), resolving all MotionElements’ migration challenges. Meilisearch offered the scalability, speed, and flexibility we needed while keeping infrastructure management minimal. Its federated search and hybrid search capabilities make it a powerful tool in MotionElements’ AI‑enhanced workflows. Federated search lets us retrieve results across multiple product types, making search in MotionElements’ one‑stop marketplace seamless. Hybrid search combines keyword search with external AI‑powered vector search, allowing us to enhance discovery across images and audio."

These features significantly improve search relevance, speed, and user experience, making it easier for the MotionElements team to deliver fast and intuitive search results.

**Implementation**
-------------------------------------------------------------------------------

Implementing Meilisearch was straightforward, thanks to its user‑friendly API. MotionElements wrote code to set up indexes based on its needs and pre‑generated NDJSON files of its products, which are stored in an S3 bucket. Meilisearch makes it easy to ingest data from NDJSON, dramatically improving MotionElements’ indexing workflow. 

Mark Sun said: _“Previously, re‑indexing on Solr took weeks, but with Meilisearch, it now takes just hours—saving us countless hours of engineering time. Since MotionElements uses the enterprise cloud version, the MotionElements developer team focuses on search optimization while Meilisearch handles the infrastructure.”_

**Results**
-------------------------------------------------------------------------------
The biggest impact has been faster search performance, improved accuracy, and higher engagement:

- **No more worrying about Solr and Zookeeper crashes**. _"Meilisearch is stable and scalable. Now, we can focus our efforts where they’re needed most, knowing our search infrastructure is stable and dependable._” — said Mark Sun.
- **Re-indexing time drastically reduced** – "what used to take weeks in Solr now takes just hours."_

- **Increased search-based conversions.** Faster, more relevant results lead to better engagement and retention.

**Conclusion: vision**
-------------------------------------------------------------------------------

Going forward, MotionElements expects further improvements in personalization, more efficient AI-powered product recommendations, and continued optimization of multilingual search, all of which are key drivers of its business growth.

As MotionElements continues investing in AI-powered search enhancements, Meilisearch remains a core search infrastructure component for the MotionElements platform. 
MotionElements’ vision includes:
- **Advancing hybrid search** – enabling seamless queries across keywords, text embeddings, image embeddings, and music similarity to deliver the best results across multiple vectors.
- **Enhancing language‑specific optimizations** – for example, in Japanese, when searching in kanji, results should also include matches in hiragana and katakana, while properly handling stop words for better accuracy.
- **Optimizing performance at scale** – indexing, updating, and searching must remain fast and efficient as MotionElements’ content library grows.
Meilisearch should continue to augment vector‑based search, rather than replace dedicated vector databases—providing a fast, flexible, and cost‑effective search infrastructure that evolves alongside MotionElements’ AI‑powered features.



Meilisearch latest news and company updates