Qdrant 1.18 - TurboQuant

2026-05-11 · Source: Qdrant Blog on Qdrant - Vector Search Engine · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Cloud Computing & IT Infrastructure · Depth: Advanced, medium

Summary

Qdrant 1.18.0 introduces TurboQuant, a new quantization method developed by Google Research, offering twice the compression ratio of scalar quantization with comparable recall and speed. This version also adds comprehensive memory monitoring via a Web UI and API, detailing disk, RAM, and page cache usage per collection component. Users can now add or remove named vectors from an existing collection schema without recreation, simplifying embedding model migrations. Further enhancements include an API endpoint for querying audit logs and support for request tracing IDs, alongside per-collection API metrics for granular performance insights. Two new strict mode guardrails, "max_resident_memory_percent" and "search_max_batchsize", enhance system stability.

Key takeaway

For AI Engineers managing vector databases, Qdrant 1.18.0 significantly improves efficiency and operational visibility. You should evaluate TurboQuant's 2x compression ratio against scalar quantization for cost savings and performance. Utilize the new memory monitoring and per-collection API metrics to diagnose performance bottlenecks and optimize resource allocation. Additionally, implement in-place named vector updates to streamline your embedding model migration workflows.

Key insights

TurboQuant offers superior vector compression and recall, while Qdrant 1.18 enhances monitoring and schema flexibility.

Principles

• Hadamard rotation normalizes vector data for efficient compression.
• In-place schema updates streamline embedding model migrations.
• Granular resource monitoring improves system diagnostics.

Method

TurboQuant applies a fast Hadamard rotation to vectors before compression, then uses length renormalization and per-coordinate calibration to optimize recall and fit data to precomputed codebooks.

In practice

• Benchmark TurboQuant for optimal performance on your data.
• Utilize new API for querying audit logs and tracing requests.
• Monitor collection memory via Web UI for capacity planning.

Topics

• Qdrant
• Vector Quantization
• TurboQuant
• Memory Monitoring
• Audit Logging
• API Metrics

Code references

• qdrant/qdrant

Best for: AI Architect, NLP Engineer, Computer Vision Engineer, AI Engineer, Machine Learning Engineer, MLOps Engineer

Related on AIssential

• Qdrant Beats Elastic’s DiskBBQ at 2x Throughput, Half the Latency, and 1/3 the Compute — Proper configuration of vector databases is crucial for optimal disk-based search performance and resource efficiency.
• The Sequence AI of the Week #834: Google's AMAZING TurboQuant for Building More Efficient AI — TurboQuant redefines AI quantization as a first-class algorithmic problem focused on high-dimensional vector geometry.
• Now Available on Qdrant Cloud: GPU Indexing, Multi-AZ, and Audit Logging — Qdrant Cloud enhances enterprise readiness with GPU-accelerated indexing, Multi-AZ replication, and comprehensive audit logging.
• Qdrant TurboQuant Explained: Is TurboQuant the Silver Bullet? — TurboQuant rotates vectors before compression, preserving geometry better for stable recall with high memory savings.
• Google Introduces TurboQuant: A New Compression Algorithm that Reduces LLM Key-Value Cache Memory by 6x and Delivers Up to 8x Speedup, All with Zero Accuracy Loss — TurboQuant offers substantial LLM KV Cache compression and speedup with no accuracy loss via data-oblivious vector quantization.
• Google unveils TurboQuant, a new AI memory compression algorithm — and yes, the internet is calling it ‘Pied Piper’ — TurboQuant offers extreme AI memory compression, reducing KV cache by 6x without performance loss.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Qdrant Blog on Qdrant - Vector Search Engine.