# Mixedbread

## Website

- [Mixedbread - Make your agents work on your data](https://www.mixedbread.com/index.md): Make your agents work on your data. Mixedbread turns PDFs, decks, videos, code, Slack, and Drive into evidence your agent can act on, with fewer tokens.
- [Pricing: plans and usage rates](https://www.mixedbread.com/pricing.md): Start free with $5 in credits, scale for $20 a month plus usage, or go enterprise. The same per-unit rates on every plan, bandwidth and egress included.
- [Quality Evals](https://www.mixedbread.com/evals.md): Toast 1 and Wholembed V3 results on public retrieval and agent benchmarks, measured against the strongest competing models.
- [ViDoRe V3 (Markdown) benchmark · Retrieval](https://www.mixedbread.com/evals/vidore-v3-text.md): Real-world, domain-specific document retrieval: ViDoRe V3 spans specialized industry corpora, evaluated on documents pre-parsed to markdown (English, French).
- [ViDoRe V3 Crosslingual (Markdown) benchmark · Retrieval](https://www.mixedbread.com/evals/vidore-v3-crosslingual-text.md): Real-world, domain-specific document retrieval in markdown, with cross-lingual queries (de, en, es, fr, it, pt).
- [ViDoRe V3 (Images) benchmark · Retrieval](https://www.mixedbread.com/evals/vidore-v3-images.md): Real-world, domain-specific document retrieval from page images rather than parsed text (English, French).
- [Miracl-Vision benchmark · Retrieval](https://www.mixedbread.com/evals/miracl-vision.md): Multilingual visual document retrieval across 18 languages.
- [Audio benchmark · Retrieval](https://www.mixedbread.com/evals/incompebench.md): Audio retrieval with graded relevance judgements.
- [Video benchmark · Retrieval](https://www.mixedbread.com/evals/video-retrieval.md): Text-to-video retrieval (general clips, localized moments, instructional).
- [BrowseComp-Plus benchmark · Agents](https://www.mixedbread.com/evals/browsecomp-plus.md): A deep-research agent answers multi-hop questions over a 100k-document web corpus.
- [MADQA benchmark · Agents](https://www.mixedbread.com/evals/madqa.md): Agents answer questions over 800 multimodal PDFs (Snowflake leaderboard).
- [OfficeQA-Pro benchmark · Agents](https://www.mixedbread.com/evals/officeqa-pro.md): OpenAI Codex answers enterprise finance questions over 89,000 pages of PDFs (Databricks benchmark).
- [BrowseComp-Plus benchmark · Toast 1](https://www.mixedbread.com/evals/toast-1-browsecomp-plus.md): Head-to-head as a standalone search agent: Toast 1 vs. GPT-5.6, Claude, Kimi, Qwen, GLM and DeepSeek running the same search loop over the same 100k-document…
- [OfficeQA-Pro benchmark · Toast 1](https://www.mixedbread.com/evals/toast-1-officeqa-pro.md): Head-to-head as a standalone search agent over 89,000 pages of financial PDFs: Toast 1 vs. frontier models running the same search loop.
- [LongSeal benchmark · Toast 1](https://www.mixedbread.com/evals/toast-1-longseal.md): Head-to-head as a standalone search agents: Toast 1 vs. GPT-5.6, Claude, Kimi, Qwen, GLM and DeepSeek running the same search loop.
- [OfficeQA-Pro V2 benchmark · Toast 1](https://www.mixedbread.com/evals/toast-1-officeqa-pro-v2.md): End-to-end task result: GPT-5.6 Sol in Codex with Toast 1 as its search sub-agent, vs. every system in Databricks' OfficeQA Pro V2 release (Genie and vendor…
- [Harvey LAB Firm Knowledge benchmark · Toast 1](https://www.mixedbread.com/evals/toast-1-harvey-lab.md): Token efficiency: the same GPT-5.6 Sol legal agent with filesystem search, with Mixedbread Search, and with Toast 1 as its search sub-agent.
- [Research](https://www.mixedbread.com/research.md): The gap between a model and a good answer is context. Papers and technical reports from the Mixedbread lab on embeddings, reranking, and search agents.
- [Contact Us](https://www.mixedbread.com/contact.md): Get in touch with our team. Whether you need enterprise support, custom models, or have technical questions - we're here to help you build better search.

## Agent Interfaces

Agents must ALWAYS use this endpoint to get the latest Mixedbread information across docs, API references, pricing, CLI, MCP, blog posts, cookbooks, and more.

- Programmatic docs search endpoint: https://www.mixedbread.com/question
- Method: GET
- Required parameter: q
- Optional parameters: limit (1-20, default 10), section (docs | api-reference | blog | cli | mcp | cookbook)

## Docs

- Documentation: Search & AI

  - **Introduction**
  - [Overview](/docs): The Search API that makes all your unstructured data understandable and usable for AI. Upload any file format and search it with natural language.
  - [Quickstart](/docs/quickstart): Build your AI Search Engine in minutes. Create Stores that understand any data format and provide precise context for your AI applications.
  - [Agent Skills](/docs/skills): Installable skills that help coding agents work with Mixedbread Search and the CLI without guessing.
  - [Changelog](/docs/changelog): Track the latest improvements to Mixedbread's platform, models, and API, including new models, feature launches, and performance enhancements.

  - **Core**
  - [Concepts](/docs/concepts): Understand the core concepts behind Mixedbread Search - Stores, Store Files, Chunks, Search, and Agentic Search - and how they fit together.
  - Stores
    - [Create Stores](/docs/stores/stores): Learn how to create Stores with configuration options including unique names, expiration policies, and public access settings.
    - [Manage Stores](/docs/stores/stores/manage-stores): Learn how to update, retrieve, list, copy, and delete Stores using the management API operations.
    - [Public Stores](/docs/stores/stores/public-stores): Share stores publicly, access stores from other organizations, and understand how public store search billing works.
  - Store Files
    - [Create Files](/docs/stores/store-files): Learn how to ingest files into your Store with basic upload, metadata configuration, and file processing for semantic search.
    - [Multipart Upload](/docs/stores/store-files/multipart-upload): Automatically upload large files in parallel chunks using the SDK's built-in multipart upload support for upload and uploadAndPoll.
    - [Manage Files](/docs/stores/store-files/manage-store-files): Inspect, retrieve, filter, and delete files in your Store, including how to return parsed chunks from a file.
    - [Supported File Types](/docs/stores/store-files/file-types): File formats supported by Mixedbread Stores for ingestion and search.
    - [Mixedbread JSON Format](/docs/stores/store-files/mxjson): Pre-chunked content format for direct ingestion into Mixedbread Stores without parsing overhead.
    - [Supported Metadata Types](/docs/stores/store-files/metadata-types): Learn about supported metadata types and how to structure metadata for optimal search performance, filtering capabilities, and content organization in Mixedbread Stores.
    - [Generated Metadata](/docs/stores/store-files/generated-metadata): Learn about the metadata generated by Mixedbread Stores for ingestion and search.
  - Search
    - [Search](/docs/stores/search): Learn how to search your Store with semantic queries, configuration options, and advanced filtering capabilities.
    - [Question Answering](/docs/stores/search/question-answering): Get AI-powered answers to your questions using Store content as context, with optional citations and multimodal support.
    - [Agentic Search with Toast 1](/docs/stores/search/agentic-search): Multi-round search driven by Toast 1 that decomposes complex questions, runs follow-up queries, and ranks the merged results.
    - [Reranking](/docs/stores/search/rerank): Learn how to use reranking to improve search result quality and relevance in your Store searches.
    - [Metadata Facets](/docs/stores/search/facets): Learn how to use metadata facets to group search results by metadata values.
    - [Web Store](/docs/stores/search/web-store): Search the web using the same API as your stores. Combine web results with your own data for comprehensive search.
  - [Metadata Filtering](/docs/stores/metadata-filtering): Learn how to filter Store files and search results using powerful metadata queries with logical and comparison operators.
  - [Data Models](/docs/stores/data-models): Understand the core data structures in Mixedbread Stores - Store Files and Chunks - and how they relate to your workflows.

  - **Agent**
  - [Responses API](/docs/agent/responses): Build agentic workflows with Mixedbread models and the OpenAI-compatible Responses API.
  - [Chat Completions](/docs/agent/chat-completions): Send messages to Mixedbread models with an OpenAI-compatible API.
  - [Hosted Tools](/docs/agent/hosted-tools): Let the server run the search loop over your Stores from the Responses API or Chat Completions.
  - [Models](/docs/agent/models): Models available through the Mixedbread Agent APIs.
  - Integrations
    - [OpenCode](/docs/agent/integrations/opencode): Use Toast 1 as a primary model or subagent in OpenCode.
  - [Build Your Own Harness](/docs/agent/build-your-own-harness): Recommendations for building custom Toast 1 search harnesses with the Chat Completions API.

  - **Production**
  - [Security & Compliance](/docs/production/security): How Mixedbread protects your data. SOC 2 Type 2 and ISO 27001 certified, GDPR compliant, with encryption everywhere and the option to keep content in storage you own.
  - [Bring Your Own Bucket](/docs/production/bring-your-own-bucket): Keep your content in object storage you own. Mixedbread indexes and searches it in memory and persists every artifact in your bucket. Nothing is retained on Mixedbread infrastructure, and every deployment comes with a downloadable Zero Data Retention attestation.
  - [Trace Export](/docs/production/tracing): Send your agent traces to Langfuse, Braintrust, or any OTLP collector you own. Every agentic search, chat completion, and Q&A run is exported as OpenTelemetry spans, the same trace the dashboard shows.

  - **Platform**
  - [Account](/docs/platform/account): All information about accounts.
  - [Organization](/docs/platform/organization): Learn about organizations in Mixedbread, including how to create them, manage members, and understand their role in resource management and team collaboration.
  - [API Keys](/docs/platform/api-keys): Create Mixedbread API keys and restrict what they can do with scopes for Stores and the completions APIs.
  - [Billing](/docs/platform/billing): Information related to billing
  - [Limits](/docs/platform/limits): Current Mixedbread store and file limits.
- API Reference: Mixedbread's API reference

  - **Mixedbread API**
  - [Introduction](/api-reference): The Mixedbread API enables powerful semantic search and document intelligence capabilities for AI-powered applications.
  - [Rate Limiting](/api-reference/rate-limits): Learn about API rate limits based on operation types. Understand token bucket policies, burst capacity, and how to handle rate limiting responses effectively.
  - [Pagination](/api-reference/pagination): Understanding cursor-based pagination across Mixedbread API endpoints. Learn how to navigate through large result sets efficiently using cursors, handle pagination parameters, and implement robust pagination logic.
  - [Error Handling](/api-reference/error-handling): Documentation on error handling with the Mixedbread API. Learn about common status codes, how to handle them, and best practices for error handling.

  - **Stores**
  - [List Stores](/api-reference/endpoints/stores/manage/list-stores): List all stores with optional search.
  - [Get Store](/api-reference/endpoints/stores/manage/get-store): Get a store by ID or name.
  - [Create Store](/api-reference/endpoints/stores/manage/create-store): Create a new store.
  - [Update Store](/api-reference/endpoints/stores/manage/update-store): Update a store by ID or name.
  - [Copy Store](/api-reference/endpoints/stores/manage/copy-store): Copy a store into a new store without re-parsing or re-embedding anything.
  - [Delete Store](/api-reference/endpoints/stores/manage/delete-store): Delete a store by ID or name.

  - **Store Files**
  - [List Store Files](/api-reference/endpoints/stores/files/list-store-files): List files indexed in a store with pagination and metadata filter.
  - [Get Store File](/api-reference/endpoints/stores/files/get-store-file): Get details of a specific file in a store.
  - [Add File to Store](/api-reference/endpoints/stores/files/add-store-file): Upload a new file to a store for indexing.
  - [Update Store File](/api-reference/endpoints/stores/files/update-store-file): Update metadata on a file within a store.
  - [Delete Store File](/api-reference/endpoints/stores/files/delete-store-file): Delete a file from a store.

  - **Store Events**
  - [List Store Events](/api-reference/endpoints/stores/events/list-store-events): List ingestion, search, grep, question answering, and chat completion events for a store with pagination and time filters.

  - **Store Search**
  - [Search Chunks](/api-reference/endpoints/stores/search/search-chunks): Perform semantic search across store chunks.  This endpoint searches through store chunks using semantic similarity matching. It supports complex search queries with filters and returns relevance-scored results.
  - [List Store Chunks](/api-reference/endpoints/stores/search/list-store-chunks): List store chunks purely by metadata filters, with no embeddings, semantic similarity or reranking. Useful for ranked retrieval over numeric metadata attributes.
  - [Question Answering](/api-reference/endpoints/stores/search/question-answering): Question answering
  - [Grep Store Chunks](/api-reference/endpoints/stores/search/grep-store-chunks): Match store chunks against a regular expression. Unlike `/stores/search`, this runs your regex against the literal text of each chunk. Use it to find chunks containing a specific token, identifier, error code, or literal phrase.
  - [Get Metadata Facets](/api-reference/endpoints/stores/search/get-metadata-facets): Get metadata facets

  - **Files**
  - [List Files](/api-reference/endpoints/files/list-files): List all files for the authenticated user.
  - [Get File](/api-reference/endpoints/files/get-file): Retrieve details of a specific file by its ID.
  - [Download File](/api-reference/endpoints/files/download-file): Download a specific file by its ID.
  - [Upload File](/api-reference/endpoints/files/upload-file): Upload a new file.
  - [Update File](/api-reference/endpoints/files/update-file): Update the details of a specific file.
  - [Delete File](/api-reference/endpoints/files/delete-file): Delete a specific file by its ID.

  - **Multipart Uploads**
  - [Create Multipart Upload](/api-reference/endpoints/files/create-multipart-upload): Initiate a multipart upload and receive presigned URLs for uploading parts directly to storage.
  - [List Multipart Uploads](/api-reference/endpoints/files/list-multipart-uploads): List all in-progress multipart uploads for the authenticated organization.
  - [Get Multipart Upload](/api-reference/endpoints/files/get-multipart-upload): Get a multipart upload's details with fresh presigned URLs for any parts not yet uploaded.
  - [Complete Multipart Upload](/api-reference/endpoints/files/complete-multipart-upload): Complete a multipart upload after all parts have been uploaded. Creates the file object and returns it.
  - [Abort Multipart Upload](/api-reference/endpoints/files/abort-multipart-upload): Abort a multipart upload and clean up any uploaded parts.

  - **Responses**
  - [Create a Response](/api-reference/endpoints/responses/create-response): Create an OpenAI-compatible response with text input or client-executed function tools.

  - **Chat**
  - [Create a Chat Completion](/api-reference/endpoints/chat/create-chat-completion): Create an OpenAI-compatible chat completion with optional client-executed function tools.

  - **Events**
  - [List Events](/api-reference/endpoints/events/list-events): List your organization's search, grep, ingestion, question answering, and chat completion events across all stores, with pagination and time filters.

  - **Integrations**
  - [Vercel](/api-reference/integrations/vercel): The Mixedbread Vercel integration makes it easy to add our powerful Search API to your Vercel projects.

  - **SDKs**
  - [Python](/api-reference/sdks/python): Learn how to install and configure Mixedbread's Python SDK for interacting with our API services.
  - [TypeScript](/api-reference/sdks/typescript): Learn how to install and configure Mixedbread's TypeScript SDK for interacting with our API services.
- CLI: Command Line Interface

  - **Getting Started**
  - [Introduction](/cli): The mxbai CLI provides a powerful command-line interface for managing Mixedbread stores and files directly from your terminal.
  - [Installation & Setup](/cli/installation): Install the mxbai CLI on your system and get started with Mixedbread's AI services.

  - **Commands**
  - [Configuration](/cli/configuration): Manage CLI settings, defaults, and aliases to customize your mxbai experience.
  - [Store Management](/cli/stores): Create, list, get, update, and delete stores.
  - [File Management](/cli/files): Upload files and manage files within stores.
  - [Search](/cli/search): Search through store content using natural language queries.
  - [Question Answering](/cli/qa): Ask questions about content in your stores and get AI-powered answers.
  - [Sync](/cli/sync): Sync files with intelligent change detection using Git-based analysis.
- MCP: Model Context Protocol

  - **Getting Started**
  - [Introduction](/mcp): The Mixedbread MCP Server connects AI assistants to the Mixedbread Search API.

  - **Reference**
  - [Tools](/mcp/tools): Complete reference for all tools available in the Mixedbread MCP server, organized by category with examples and best practices.

  - **Integration Guides**
  - [Claude Desktop](/mcp/integrations/claude-desktop): Set up the Mixedbread MCP server with Claude Desktop App.
  - [Claude Code](/mcp/integrations/claude-code): Set up the Mixedbread MCP server with Claude Code CLI.
  - [Cursor](/mcp/integrations/cursor): Set up the Mixedbread MCP server with Cursor IDE.
  - [Codex CLI](/mcp/integrations/codex-cli): Set up the Mixedbread MCP server with the Codex CLI.
  - [Gemini CLI](/mcp/integrations/gemini-cli): Set up the Mixedbread MCP server with the Gemini CLI.
  - [OpenCode](/mcp/integrations/opencode): Set up the Mixedbread MCP server with OpenCode.
  - [Zed](/mcp/integrations/zed): Set up the Mixedbread MCP server with Zed.
  - [Windsurf](/mcp/integrations/windsurf): Set up the Mixedbread MCP server with Windsurf.
  - [VS Code](/mcp/integrations/vs-code): Set up the Mixedbread MCP server with VS Code.
- Cookbook: Practical guides & examples
  - [Cookbook](/cookbook): Practical guides and examples for building with Mixedbread's AI search and embedding capabilities.

  - **Guides**
  - [Chat with PDFs](/cookbook/chat-with-pdfs): Build a document Q&A system with PDFs, slides, and rich media using Mixedbread's native search and question answering.
  - [Text-to-Image Search](/cookbook/image-search): Build a semantic image search engine using Mixedbread's multimodal and multilingual capabilities.
  - [Build an AI Agent with Memory](/cookbook/agent-memory): Build AI agents with infinite, searchable memory using Mixedbread Stores as a semantic file system.

## Blog

- [Asymmetric Quantization: Near-Lossless Late Interaction Retrieval with 97% Storage Reduction](/blog/asymmetric-quant): Late interaction makes retrieval more precise but turns every document into many vectors. Asymmetric quantization keeps query vectors precise and stores document vectors as binary signs, cutting corpus storage 32x while holding 89.65 vs 90.26 NDCG@10.
- [64 bytes per embedding, yee-haw 🤠](/blog/binary-mrl): Binary MRL combines two popular approaches to deal with the scalability issues of embeddings. It helps our embedding model achieve a 64x gain in efficiency while retaining more than 90% of performance, drastically reducing infrastructure costs and enabling new applications.
- [Closing the Oracle Gap for Your Agents](/blog/closing-gap): Mixedbread Search v3 narrows the oracle gap for agentic retrieval, topping BrowseComp-Plus and setting leading results on MADQA and OfficeQA-Pro.
- [Open Source Gets DE-licious: Mixedbread x deepset German/English Embeddings](/blog/deepset-mxbai-embed-de-large-v1): Introducing deepset-mxbai-embed-large-v1, a new open-source German/English embedding model, developed through collaboration between deepset and Mixedbread. This model sets a new performance standard among open source peers, supporting binary quantization and Matryoshka representation learning for significant cost reductions. Outperforming domain-specific alternatives in real-world applications, it offers 97%+ infrastructure cost savings through binary MRL.
- [Baking in Performance - Dynamic Batching with Batched](/blog/dynamic-batching): Learn how dynamic batching boosts GPU efficiency in machine learning. We introduce Batched, our open-source library that makes it easy to implement dynamic batching and improve performance in your ML projects.
- [Fantastic (small) Retrievers and How to Train Them: mxbai-edge-colbert-v0](/blog/edge-v0): Introducing our new family of extremely efficient ColBERT models, to serve as backbones for modern late interaction research while outperforming ColBERTv2 with just 17 million parameters.
- [Building a Robust Ingestion System 
for Any File of Any Size](/blog/infinite-file-sizes): A 300GB video does not fit on one machine, and a 4MB PDF can render into gigabytes of images. This is how our slicer and parser architecture processes arbitrary files with bounded compute and memory.
- [Getting Better with Baguetter - New Retrieval Testing Framework](/blog/intro-baguetter): Baguetter is our new open-source retrieval experimentation framework that combines sparse, dense, and hybrid retrieval into a single, easy-to-use interface. It enables fast benchmarking, implementation, and testing of new search methods.
- [BM𝒳: A Freshly Baked Take on BM25](/blog/intro-bmx): Introducing BMX, an iteration on the industry standard BM25 search algorithm. Through the incorporation of entropy-weighted query-document similarity and weighted query augmentation, the algorithm can increase search performance on the most relevant information retrieval benchmarks.
- [Dense Retrievers Know More Than They Can Express](/blog/latent-terms): Demonstrating that dense retrieval models learn much more information than they can express through their usual scoring mechanism: they also contain an indexable, natural-language-like sparse vocabulary, which is plug-and-play with BM25.
- [Ranking Beyond Binary Relevance: mxbai-rerank-v3-listwise](/blog/listwise-rerank): Announcing mxbai-rerank-v3-listwise, our new listwise reranker codesigned with Wholembed v3. It improves results on every benchmark we ran, with state-of-the-art instruction following.
- [maxsim-cpu: Maximising Maxsim Efficiency](/blog/maxsim-cpu): Introducing maxsim-cpu, a much faster way to compute the late interaction's MaxSim operator on modern CPU hardware, optimised for both x86 and Mac ARM.
- [Introducing Mixedbread Search](/blog/mixedbread-search): Introducing Mixedbread Search, the first search API built from the ground up for the AI era, with both humans and AI in mind. Natively multi-modal and multi-lingual, Mixedbread Search eliminates all friction and lets you find information where it lives.
- [Inside Mixedbread: 
How We Built Multimodal Late-Interaction at Billion Scale](/blog/multimodal-late-interaction-billion-scale): Technical deep-dive into Mixedbread Search - the first production-ready late-interaction search with native multimodality. Learn how we achieve ~80ms end-to-end latency on billion-scale document collections.
- [ColBERTus Maximus - Introducing mxbai-colbert-large-v1](/blog/mxbai-colbert-large-v1): mxbai-colbert-large-v1 is a state-of-the-art ColBERT model for reranking and retrieval tasks. It is based on the mxbai-embed-large-v1 model and achieves state-of-the-art performance on 13 publicly available BEIR benchmarks. It's available on Hugging Face.
- [Fresh 2D-Matryoshka Embedding Model](/blog/mxbai-embed-2d-large-v1): The 2D-🪆 model introduces a novel approach that enables you to reduce both the number of layers and the dimensions of embeddings within the model. This dual reduction strategy allows for a more compact model size while still delivering competitive performance compared to leading models, such as Nomic's embedding model. Specifically, reducing the model's layers by approximately 50% retains up to 85% of its original performance, even without additional training.
- [Open Source Strikes Bread - New Fluffy Embedding Model](/blog/mxbai-embed-large-v1): Our English embedding model provides state-of-the-art performance among other efficiently sized models. It outperforms closed source models like OpenAI's text-embedding-v3.
- [Every Byte Matters: Introducing mxbai-embed-xsmall-v1](/blog/mxbai-embed-xsmall-v1): Announcing mxbai-embed-xsmall-v1, our smallest and most efficient English embedding model optimised for retrieval tasks. It comes with competitive performance at an extra small footprint, support for long context, binary quantization and Matryoshka representation learning.
- [Boost Your Search With The Crispy Mixedbread Rerank Models](/blog/mxbai-rerank-v1): Introducing Mixedbread rerank models - Upgrade your search results with our new, open-source reranking models from Mixedbread. These models, available in three sizes, make it easier to find relevant results by adding a semantic layer to existing search systems. They're simple to use, work with your current setup, and are proven to boost performance with many traditional and semantic search models. Check them out for a more accurate, efficient search experience.
- [Baked-in Brilliance: Reranking Meets RL with mxbai-rerank-v2](/blog/mxbai-rerank-v2): Introducing mxbai-rerank-v2, the second-generation reranking models from Mixedbread. They're crispier than ever—featuring reinforcement learning, multilingual support, and extended context handling for even better accuracy. They come with the same open-source flexibility under Apache 2.0, now with boosted performance for a more powerful search experience.
- [mxbai-rerank-v3.1-listwise](/blog/mxbai-rerank-v3-1-listwise): Meet listwise v3.1, a small upgrade to listwise v3. Now in Mixedbread Search, matching gpt-5.6-sol (high) quality at 61x lower latency.
- [How PlanetScale Powers the Mixedbread Control Plane](/blog/planetscale-case-study): Why Mixedbread moved to PlanetScale for database reliability, low latency, observability, and hands-on support.
- [A Delicious Free Lunch: Better Projections Improve ColBERT](/blog/projection-variants): Discussing the unique learning constraints introduced by the MaxSim operator, and demonstrating that simple architecture improvements to accommodate for these limitations can increase performance in a free-lunch fashion.
- [Our Research Vision, Part 1](/blog/research-vision): Introducing our research vision and our ultimate goal: Solving Search. In this blog post, we lay out our view on Mixedbread as a research lab, our approach for prioritizing our research and making it impactful for our journey toward the objective, as well as our thoughts on balancing open- and closed-source research.
- [The Hidden Ceiling: How OCR Quality Limits RAG Performance](/blog/the-hidden-ceiling): Benchmarking shows OCR errors cap text-based RAG: top OCR still misses 4-5% NDCG@5, while Mixedbread's multimodal vector store beats perfect text by 12% and recovers 70% of lost answer accuracy.
- [Introducing Toast 1](/blog/toast-1): Meet Toast 1, Mixedbread's search agent for knowledge-intensive tasks, matching or outperforming Claude Opus 5 and GPT-5.6 Sol while being up to 10× cheaper and 12× faster.
- [Mixedbread now live on the Vercel Marketplace](/blog/vercel-marketplace): Announcing the integration of Mixedbread Search with the Vercel Marketplace. Seamlessly integrate state-of-the-art, blazingly fast, multi-modal semantic search into your Vercel projects!
- [Beyond the Limit: Introduce Mixedbread Wholembed v3](/blog/wholembed-v3): Announcing Mixedbread Wholembed v3, our new unified omnimodal multilingual retrieval model, setting a new state of the art for search across languages, modalities, and real-world retrieval tasks.