1. X
  2. Mixedbread
Log inSign up
Mixedbread
136 posts
Image
user avatar
Mixedbread
@mixedbreadai
The Retrieval Infrastructure for agents.
San Francisco, CA
mixedbread.com
Joined March 2024
9
Following
3,453
Followers
AffiliatesAffiliatesRepliesRepliesArticlesArticlesMediaMedia

New to X?

Sign up now to get your own personalized timeline!

Create account

By signing up, you agree to the Terms of Service and Privacy Policy, including Cookie Use.

Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
Don't miss what's happening
People on X are the first to know.
Log inSign up
  • Pinned
    user avatar
    Mixedbread
    @mixedbreadai
    Mar 12
    Introducing Mixedbread Wholembed v3, our new SOTA retrieval model across all modalities and 100+ languages. Wholembed v3 brings best-in-class search to text, audio, images, PDFs, videos... You can now get the best retrieval performance on your data, no matter its format.
    Image
    204K0204K
  • user avatar
    Mixedbread
    @mixedbreadai
    Jul 15
    From a 1 MB document to a 300 GB video, there is always a bigger file. Mixedbread indexes text, PDFs, images, videos for your agents to retrieve the right evidence. We built the system to let you ingest any file of any size.
    Image
    GIF
    12K012K
    user avatar
    Mixedbread
    @mixedbreadai
    Jul 15
    Replying to @mixedbreadai
    More lessons from building a robust ingestion pipeline: - stream video slices with ffmpeg from presigned URLs - adjust PDF DPI to fit large page geometries into budget - yield chunks lazily, one at a time - start new slices at the last complete chunk boundary
    Image
    GIF
    4820482
    user avatar
    Mixedbread
    @mixedbreadai
    Jul 15
    More about how we rebuilt our file ingestion system:
    Building a Robust Ingestion System for Any File of Any Size
    Building a Robust Ingestion System for Any File of Any Size
    From mixedbread.com
    2540254
  • user avatar
    Mixedbread
    @mixedbreadai
    Jun 29
    Article cover image
    Article
    Asymmetric Quantization: Near-Lossless Late interaction Retrieval with 97% Storage Reduction
    TL;DR: We built a multimodal and multilingual search engine using late interaction. In order to serve it, we built our own object storage based vector database. Here we talk about asymmetric...
    22K022K
  • user avatar
    Mixedbread
    @mixedbreadai
    Jun 26
    New: Granular evidence citations in QA Beyond chunk-level citations, our QA now points to the exact evidence snippet supporting each answer. Easier to inspect, verify, and trust. → Available with high-quality processing.
    Image
    00:00
    1.8K01.8K
    user avatar
    Mixedbread
    @mixedbreadai
    Jun 26
    Our high-quality processing detects the layout of documents, further supporting: - granular QA citations - your downstream LLM understanding
    Image
    9720972
    user avatar
    Mixedbread
    @mixedbreadai
    Jun 26
    Build your own granular citation with Mixedbread:
    Image
    2650265
  • Mixedbread reposted
    user avatar
    Alexander Martin
    @alexdmartin314
    Jun 24
    @mixedbreadai's wholembed-v3 set a high bar for video retrieval as a single-stage retriever on our MAGMaR shared task! To beat it, both C2F-RAG and MARQUIS needed multi-stage pipelines with reasoning-based reranking on top. That's a strong model.
    Image
    user avatar
    MAGMaR
    @MAGMaR_workshop
    Jun 23
    Replying to @MAGMaR_workshop
    The top 2 systems are C2F-RAG by Dai et al, using text captions + LLM-based cognitive reranking, and MARQUIS by @derangineer et al, using query decomposition and video native reranking.
    4.8K04.8K
  • user avatar
    Mixedbread
    @mixedbreadai
    Jun 23
    New: Bring your own bucket Mixedbread is built on top of object storage. Now that storage can be yours. Your data stays in a bucket you control. Mixedbread indexes and searches with zero content retention on our side. For enterprise teams that need control, compliance, audit
    Image
    00:00
    4.1K04.1K
    user avatar
    Mixedbread
    @mixedbreadai
    Jun 23
    Powerful search in your control:
    Image
    1.2K01.2K
  • user avatar
    Mixedbread
    @mixedbreadai
    Jun 10
    New: Metadata explorer Adding metadata to files enables filtering during search. Now, you can browse metadata fields and values across your store.
    Image
    00:00
    1.9K01.9K
    user avatar
    Mixedbread
    @mixedbreadai
    Jun 10
    Agents can inspect file metadata in a store to understand available filters. docs: mixedbread.com/docs/stores/se…
    Image
    3160316
  • user avatar
    Mixedbread
    @mixedbreadai
    Jun 2
    By now, everyone knows that single-vector embedding models are hugely limiting for modern workflows. But they contain than you think: you can extract sparse Latent Terms from them. And it turns out that BM25 is all you need to turn this vocabulary into a strong retriever.
    Image
    00:00
    41K041K
    user avatar
    Mixedbread
    @mixedbreadai
    Jun 2
    Replying to @mixedbreadai
    Having language-adjacent properties means that tools designed for lexical approaches "just work". BM25, always refusing to exit the scene, is strong here: applied over the Latent Terms extracted from nomic-embed-v1.5, it results in a near state-of-the-art sparse retriever.
    Image
    2.7K02.7K
    user avatar
    Mixedbread
    @mixedbreadai
    Jun 2
    Read more here:
    Dense Retrievers Know More Than They Can Express
    Dense Retrievers Know More Than They Can Express
    From mixedbread.com
    1.5K01.5K
  • user avatar
    Mixedbread
    @mixedbreadai
    May 27
    New: grep for exact matching grep → keyword / regex matching search → fine-grained semantic retrieval Works across uploaded content, including text, PDFs (OCR) and audio/video (transcription). Give your agents both retrieval primitives to perform at their best.
    Image
    5.6K05.6K
    user avatar
    Mixedbread
    @mixedbreadai
    May 27
    docs:
    mixedbread.com
    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,...
    6510651
Advertisement
Advertisement