Qdrant

High-performance Rust-based vector search engine.

[Qdrant @qdrant_engine]

Already running OCR? The recently released Rerank 5 Pro lifted nDCG@10 on the same corpora from 66.8 to 73.2.

Full breakdown by @krotenWanderung: https://t.co/1bhnKuqLeC

[Qdrant @qdrant_engine]

Can one dense vector per PDF page compete with late interaction?

@cohere's Embed 5 Pro turns a whole page image into a single 2048-dimensional vector, so we tested it against models like ColPali.

[Qdrant @qdrant_engine]

On five public ViDoRe V3 corpora, Embed 5 Pro scored 65.8 nDCG@10, ahead of ColPali, ColQwen2.5, and Jina-v4. Only ColEmbed-3B-v2 scored higher (67.5) of those we compared. Embed 5's training data isn't published, so test set overlap is possible.

[Qdrant @qdrant_engine]

Qdrant 1.20: Ship Hybrid Search without the sparse pipeline. Turn on BM25 scoring for a full-text index, query it with your vectors in one request. Plus a code tokenizer, faster nested filters, turbo16, and turbo8. https://t.co/y5rYFwDkU3

[Qdrant @qdrant_engine]

Vector search ranks results. But who counts the filters?

Parichay Das shows how Qdrant's Facet API, payload indexes, and disjunctive faceting power accurate search sidebars.

Including a key trade-off: filtered counts ≠ relevant results.

https://t.co/PpIzK6SdhB

[Qdrant @qdrant_engine]

Better chunking ≠ better retrieval.

At MLOps Community Amsterdam, Qdrant's @EASzyszka explored chunking benchmarks, while https://t.co/kxmr3XmSYJ shared lessons on agent memory and source consistency.

Luca Berton's recap:

https://t.co/EyBC16ixK3

[Qdrant @qdrant_engine]

London Search Week, let's talk search over a pub quiz.

Join Qdrant on Nov 24, 6:30–9:30 PM GMT at Core Bar, London.

Search trivia, prizes, food & your first drink on us. Come solo or bring a team of up to 6.

Limited spots. RSVP:

https://t.co/nZVCqjkNvx

[Qdrant @qdrant_engine]

Your vector collection outgrew RAM. What now?

Qdrant 1.19 lets you choose pinned, cached, or cold memory tiers per component.

Keep original vectors on disk and compressed vectors in RAM to balance memory use and search latency.

https://t.co/eNGYfL9c13

[Qdrant @qdrant_engine]

How is AI changing software development?

Qdrant's @EASzyszka joined @WeAreDevs at Agent Conf Berlin to discuss token usage, vector search, data quality, and the evolving developer workflow.

Watch the conversation:

https://t.co/moUO0aHT1g

[Qdrant @qdrant_engine]

Which German councils are discussing wind power?

Poliscope answers in one search.

4,800 council portals, 130M points, one self-hosted Qdrant node.

Searches went from up to 20 s to 0.87 s on average with hybrid search, tenant filters, and TurboQuant.

https://t.co/XgYfVeAm6G

[tomaarsen @tomaarsen]

The fine folks at @qdrant_engine are doing great work with Qdrant Edge to make vector search accessible on edge devices, such as the @huggingface Reachy Mini!

Great for memories, as they show below! Check it out.

[Bo @bo_wangbo]

very happy to see Qdrant team testing our latest late interaction retriever!

[Qdrant @qdrant_engine]

https://t.co/uXP12n7r2A

[Qdrant @qdrant_engine]

21× smaller vectors. 99.9% of baseline retrieval quality retained with rescoring.

We tested Perplexity’s pplx-embed-v2-late with Qdrant’s TurboQuant across 4 ViDoRe v3 datasets.

856 KB → 40 KB per page, with a trade-off: rescoring requires the original vectors.

⁠Full benchmarks below 👇

[Qdrant @qdrant_engine]

@OptimistOtaku Great work!!

[Qdrant @qdrant_engine]

Read the benchmarks: https://t.co/ZlnUTTHGYw

[Qdrant @qdrant_engine]

780 ms during background optimization. 4.3 ms after it finished.

Same Qdrant collection, very different search latency.

We tested 13 optimizer configurations on 1.76M vectors to understand the trade-offs between indexing, recovery time, and query performance.

Here’s what we found 👇