Redis 8.0

What's new in Redis 8

Redis 8.0 introduces powerful new capabilities, including the beta release of the Vector Set data structure, designed for AI use cases such as semantic search and recommendation systems.

This release also delivers over 30 performance improvements, including:

• Up to 87% faster command execution
• Up to 2x higher throughput
• Up to 18% faster replication

Additionally, the Redis Community Edition has been renamed to Redis Open Source, marking a shift toward a unified, modern distribution. This version offers the full power of Redis — no modules, no fragmentation — just Redis.

Below is a detailed breakdown of these updates.

New features

Vector set [beta]

The new Vector set data structure supports high-dimensional vector similarity search, which is ideal for AI use cases such as semantic search and recommendation systems. Vector set complements Redis’s existing vector search capabilities via the Query Engine and is currently available in beta. APIs and behaviors may change in future releases.

New hash commands

Redis 8 builds on Redis 7.4’s introduction of field-level expiration in hashes and adds three new commands for working with hashes:

• HGETEX: Fetch a hash field and optionally set an expiration
• HSETEX: Set a hash field and optionally set an expiration
• HGETDEL: Fetch and delete a hash field

These commands simplify common caching and session management usage patterns.

Improvements

Redis Query Engine Improvements

The Redis Query Engine allows users to use Redis as a document database, a vector database, a secondary index, and a search engine. With Redis Query Engine, users can define indexes for hash and JSON documents, and use a rich query language for vector search, full-text search, geospatial queries, and aggregations.

Use cases include:

• Vector Search: Search based on semantic similarity using vector embeddings stored in JSON or hashes
• Exact Matching: Lookups using filters, tags, and ranges
• Full-Text Search: Supports stemming, synonyms, and fuzzy matching

The Redis Query Engine powers a wide range of applications, from AI retrieval-augmented generation (RAG) systems to full-featured search engines.

Access Control Improvements

Access Control Lists (ACLs) have been updated to support the new data structures introduced in Redis 8. Existing ACL categories such as @read and @write now include commands for JSON, time series, VECTOR, and probabilistic data structures.

These updates allow for more precise control over which operations users can perform on each data structure.

Performance improvements

Redis 8 delivers the largest performance leap in Redis history with over 30 optimizations, including:

• Up to 87% lower command latency
• 2x throughput in multi-threaded workloads
• 35% memory savings for replica nodes
• 16x more query processing capacity with horizontal and vertical scaling

These improvements benefit both single-node and clustered deployments. More details are available in the Redis 8 GA blog post.

Changes

Breaking changes

ACL behavior

Commands from included modules are now covered under standard categories (e.g., +@read, +@write). A user with +@all -@write will no longer be able to execute FT.SEARCH as they could before. Explicit inclusion of new command categories is required to maintain access.

Redis Query Engine

The following changes affect behavior and validation in the Redis Query Engine:

• Enforces validation for LIMIT arguments (offset must be 0 if limit is 0).
• Enforces parsing rules for FT.CURSOR READ and FT.ALIASADD.
• Parentheses are now required for exponentiation precedence in APPLY expressions.
• Invalid input now returns errors instead of empty results.
• Default values revisited for reducers like AVG, COUNT, SUM, STDDEV, QUANTILE, and others.
• Updates to scoring (BM25 is now the default instead of TF-IDF).
• Improved handling of expired records, memory constraints, and malformed fields.

For a full list of the Redis Query Engine-related changes, see the release notes.

Other changes

One Redis

Over time, Redis introduced several modules to address new use cases such as search, vector similarity, time series analysis, and probabilistic modeling. While these modules extended Redis’s functionality, managing version compatibility and module installation introduced friction.

Redis Stack partially solved this by bundling modules together, but it also created fragmentation across the ecosystem.

Redis 8 merges Redis Stack and Redis Community Edition into a single unified distribution: Redis Open Source. All previously modular functionality is now built into the Redis Open Source package, eliminating the need to manage separate modules. You now get a single, consistent feature set across all deployments.

Component versions

Redis 8.0 delivers just Redis — no modules, no fragmentation.