Mixedbread
Mixedbread is an applied research lab and retrieval-infrastructure company building the memory for AI. Its API and search stack ingest multimodal, multilingual data and return precise, auditable evidence for AI agents and applications.
At a Glance
- AI application and agent developers
- Enterprise software and knowledge-management teams
- Research and information-retrieval teams
- Companies with multimodal or multilingual data
- +2 more
AI Tools by Mixedbread
(2)
mxbai-rerank
Open Source Reranking Models for RAG
Mixedbread
Multimodal Search API for AI
Discussions
No discussions yet
Be the first to start a discussion about Mixedbread
Latest News
How PlanetScale Powers the Mixedbread Control Plane
Introducing Toast 1, Mixedbread’s specialized search agent
mxbai-rerank-v3.1-listwise released as the default reranker
Building a robust ingestion system for any file of any size
Products & Services
Managed multimodal, multilingual search API: upload PDFs, images, documents, code, audio and video into Stores, automatically process them, and retrieve AI-ready evidence with natural-language search, semantic search, reranking and agentic search.
Specialized search agent that decomposes queries, gathers evidence, inspects sources and curates context; can run standalone or as a subagent.
Default listwise reranker designed for complex instructions, recency-aware ranking and source-priority resolution.
Listwise reranking model co-designed with Wholembed v3, with natural-language instruction following.
Market Position
Mixedbread positions itself as a research-led, end-to-end alternative to assembled retrieval stacks: it trains its own retrieval models, operates the search/indexing infrastructure and exposes a managed multimodal API. It competes with vector-database/RAG platforms and managed search providers such as Pinecone, Weaviate, Elasticsearch/OpenSearch, Vespa, Cohere Rerank and Voyage AI, differentiating on native multimodality, late-interaction retrieval, multilingual coverage, agentic search, evidence provenance and low latency.
Leadership
Founders
Aamir Shakir
Co-founder; studied at EPFL (École polytechnique fédérale de Lausanne) and has worked on information retrieval, embeddings and reranking research at Mixedbread.
Julius Lipp
Co-founder and former CTO of Mixedbread (October 2023–June 2025); software developer with prior experience at Google.
Executive Team
Aamir Shakir
Co-Founder
AI research and retrieval expert.
Julius Lipp
Co-Founder
AI and search systems engineer.
Board of Directors
Founding Story
Mixedbread was set up as a research lab to pursue ‘Perfect Search’: the founders believed information retrieval is a cornerstone of intelligence and that useful research should feed directly into a real product. The initial vision was to build a unified engine that gives AI systems the right context across the world’s knowledge, rather than expecting models to memorize changing, specialized information.
Business Model
Revenue Model
Usage-based managed API/search infrastructure: customers pay for indexing, search queries, storage and Toast 1 LLM-token usage; paid plans add credits, higher limits, backups and support. Enterprise contracts add volume discounts and dedicated infrastructure/support.
Pricing Tiers
$5 one-time credits; 3 workspace users; 3 stores; 100 queries/min and 6,000 uploads/min; community Slack support.
$20 monthly credits then pay-as-you-go; unlimited users; 10,000 stores; 2,400 queries/min and 12,000 uploads/min; automatic backups and priority Slack support.
Volume discounts, unlimited users/stores, custom rate limits and SLA, automatic backups, point-in-time recovery, BYOB/BYOC and dedicated support.
Target Markets
- AI application and agent developers
- Enterprise software and knowledge-management teams
- Research and information-retrieval teams
- Companies with multimodal or multilingual data
- Developers needing managed RAG/search infrastructure
- Startups and venture-backed companies (startup credits program)
- Retrieval-augmented generation and AI assistants
- Knowledge-intensive research and question answering over enterprise corpora
- Coding agents and search over code/SQL repositories
- Search across PDFs, decks, tables, images, audio and video
- Enterprise knowledge bases connected to Slack and Google Drive
- Auditable agent workflows requiring citations and evidence provenance
- Vercel
- AnkiHub
- TinyFish
- Prime Intellect