Explore tweets tagged as #mixedbread
We build the first production ready multi-vector and multimodal search. Now we are serving over 1 billion documents in under 50ms latency (p50). We are sharing how we build it.
15
43
329
尝试了 mixedbread 的 mgrep，真的让 AI 搜索时的效率变的非常高，既准确又快速。而且还可以节省 AI Token 的消耗。 然后买了 $20/月的 Scale 计划。在团队内推广了下。 6 天后收到了用量限制邮件，点开一看，4 个人6 天花费了 $50... 节省的 Token 开销又成倍的回来了🥲
1
0
1
We're building Mixedbread to close the gap between Search that is possible today and what the users of tomorrow will demand. You can read more about it here:
3
4
54
the industry is moving in an interesting direction with semantic search alternatives to grep for coding agents. there's mgrep by mixedbread there's semtools by llamaindex now there's (multi-vec) colgrep by lighton very cool to see these advancements. congrats to
Your coding agent is burning tokens on grep like it's 1973 Because semantic search means remote APIs & babysitting an index Introducing ColGrep & LateOn-Code SOTA code retrieval with lightweight models. Wins 70% vs grep. 15.7% less tokens. Local, open & free. Runs on a toaster.
7
12
123
Some updates on mixedbread rerank vs. cohere. I have been running tests of increasing complexity: 1. Synthetic dataset: Simple retrieval over 50 chunks, 25 queries Sample query: "Which act explicitly applies to all types of personal information processing?" BM25: 0.8 Cohere:
1
0
20
One More (Small) Thing: Introducing mxbai-colbert-edge-v0 17M and 32M. They are are the result of an easily reproducible way to train ColBERT models from scratch. They're strong, too: the 17M variant would rank first on the LongEmbed leaderboard for models under 1B parameters.
5
23
129
Learn how to Boost your RAG and search performance with Mixedbread’s purpose-built models for high-quality retrieval across text, code, and tools. In this upcoming Learning Together session on July 15th, Mixedbread CEO Aamir Shakir will dive into: • How reinforcement learning
1
5
26
> be me, go to SF > never did a hackathon before > find the context engineering one from @Theoryvc > claude code yolo with Mixedbread > won > have a new PC
3
6
43
An explainer of trusty ol' BM25! There was recently a BM25 extension proposed by mixedbread and I thought it'd be helpful to share a resource that clearly explains how the hyperparams in the original BM25 work! This video is an awesome explanation of b and k!
1
13
58
Mgrep —— 这玩意儿基本上就是装了 AI 大脑的 Grep，简直是代码搜索外挂。 简单说，就是个"会读心术"的文件搜索助手，让你不用猜文件名，直接说想要什么就行。 据说跑过50个任务测试，用mgrep的AI助手token消耗减半，质量还更好！ https://t.co/bjxwKKtYnu
2
42
299
Seriously cool SotA rerankers from Mixedbread! They can handle 100 languages, reranking code, technical docs, function calls, JSON, e-commerce products etc. Trained Qwen2.5 0.5B and 1.5B using a GRPO -> Contrastive -> RL recipe. 55.57 and 57.49 on BEIR for 0.5B and 1.5B
1
4
29
⚡ The new mxbai-embed-2d-large-v1 embeddings model from mixedbread is impressive! Performance with 384 dimensions using 2D Matryoshka Sentence Embeddings is superb. This model works out of the box with txtai - see below. https://t.co/CeEvRYXKOt
https://t.co/qnyvRa3Grv
1
2
8
You've never heard of Dun Zhang from Nantong University but he trained an embedding model that outperforms models from OpenAI, Cohere, Voyage, Jina, Snowflake, Nomic, and MixedBread (at least in English) Now help me parse this short Dec 26th technical report he dropped about it
4
38
469
Tuesday in Open Access ML: - IBM silently dropping Merlinite 7b - Moondream2 - small VLM for edge - TripoSR - image-to-3D (StabilityAI +Tripo) - Microsoft Orca Math dataset - Mixedbread 2D Matryoshka embeddings - Based by HazyResearch + Together 🚀
2
26
130
The raise and fall of cosine similarity 2024, featuring Cohere and Mixedbread
4
4
76
まとめると3点。①Vercel AI GatewayがGrokの動画・画像生成に対応、明日まで無料。②Vercel Workflowsで生成中のブラウザ落ちや回線切断を自動カバー。③mixedbread aiによるビジュアルベクトル検索で、テキストから画像を意味検索できるCreative
Vercel AI Gateway now supports video generation. Grok Imagine Video & Image are 🆓 until tomorrow. We used @v0 to create an open source Creative Studio powered by @xai Grok. Create images, videos, or make your own design tool! https://t.co/07QZC6bmwT – it's quite fast. Some
0
0
0
Mixedbread makes image search feel like magic. No keyword labeling, no manual annotation. Describe what you are looking for and instantly find it.
1
5
20
@tiagoefreitas Vespa, LightOn, MixedBread of course, NeuML, and recently NVIDIA have contributed so much to this space, plus varying levels of support in Qdrant and Weaviate. I don't track it as much, but HuggingFace still has ColBERTv2 at 15M downloads per month: https://t.co/znLs4Xyd7E
1
0
21