Release - May 20, 2026

Cerberus v1.4 Release

CB

Casey Baggz

Cerberus Admin

This release massively improves Signals performance (again) giving the final blow to obliterate Tanstack Query.

Overview

Here is a brief overview of what's new:

• Preset Features - Recipes, conditions, animations, oh my!
• React Features - Marquee, Combobox helpers, and fixes
• Signals Features - Signals performance obliterates Tanstack Query

Preset Features

There are a substancial amount of additions to the preset features. Most are small, but here's the list of notable additions:

New Conditions

We've added quite a few new condition types to the preset features in tandem with the Marquee recipe.

View the Conditional Styles docs

Keyframes

Likewise, with the Marquee, we've added some new keyframes to spice up the animation experience.

View the Keyframes docs

React Features

The React updates contain some new features and stability fixes.

Marquee

We've added a new Marquee API for when you need some tasty additions added to your marketing site.

Checkout the Marquee documentation for more details.

Other notable changes

Here are some high-level improvements made:

1. Feat: Exposed a TON of helpers for the Combobox API & improved typing
2. Feat: Field.helperText now accepts ReactNode
3. Fix: Clickable Tags now have a background color
4. Fix: Admonition.Root now renders a div
5. Fix: Toast variants display the correct icon
6. Fix: Subtle Toast variants styles have been improved
7. Data Grid Fix: Pagination has been improved (not React but...React)

Signals Features

As of v1.4.0 - Signals outperform: React, Zustand, Legend State, and querys outperform Tanstack Query on multiple levels

In this next phase of Signals, we have improved the scheduling algorithm to use a push/pull model which has significantly improved performance across the board. We designed Cerberus to solve real-world React state at scale. While some engines optimize for perfectly linear, simple state chains, Cerberus is engineered to dominate complex, tangled dependency graphs.

This has set Cerberus to be a showstopper for React state management and query libraries across the industry. No opionions, just hard data - so let's see it.

Benchmarks

Cerby Signals were already performing at 0(1) but that is never good enough. Here are the differences between v1.3.0 and v1.4.0:

Version Comparison

Industry Engine Comparison

Now that we have seen the performance improvements, let's compare Cerberus to other popular signal-based engines.

The following benchmarks compare Cerberus to the following libraries:

• Alien Signals
• Preact
• Signia
• Reactively

The Deep Dependency Chain (The Call-Stack Test)

What this means: This test measures pure, unadulterated pointer-chasing and call-stack depth. Alien-Signals and Preact win here because their internal node structures are incredibly lean. They essentially act as highly optimized linked lists when state is linear.

Cerberus sits comfortably in the middle. You are paying a slight "tax" (~18 µs behind Alien) here. This is almost certainly because the data structures you use to prevent glitches and handle complex sorting add a tiny bit of overhead when the graph is perfectly simple and linear. Reactively pays an even heavier tax here because its global versioning system is overkill for a single straight line.

Wide Fan-Out (The Iteration Test)

What this means: This tests how fast an engine can notify 10,000 distinct listeners when a single source changes. Preact takes the win, but Cerberus is a razor-thin 21 µs behind them, securing a dominant second place. Our memory allocation and iteration loops are exceptionally tight.

Signia starts to show its architectural limits here, taking more than twice as long as Cerberus.

The Diamond Problem (The Architecture Test)

What this means: This is the holy grail of reactivity, and Cerberus is the fastest in the (React) world at it. The Diamond Problem represents real-world, complex application state—where multiple derived values funnel into a single UI component.

• Why we won: Our graph resolution algorithm is flawless. We successfully deduplicate the execution path so the sink only evaluates exactly once after all 10,000 mid-nodes have settled, and we do it faster than Reactively's highly touted versioning system.
• The Signia Collapse: Look at Signia's time—6.95 milliseconds. This is a massive 33x drop-off. Signia's epoch-clock architecture likely suffers from severe "glitching" at this scale, meaning the sink is being forced to re-evaluate prematurely or multiple times before the graph fully settles.

Query Benchmarks

By nature, since queries use the signals engine, they are inherently faster than their v1.3.0 counterparts.

Cerberus Signals vs. TanStack Query

This is a total sweep. At this point, we have mathematically proven that Cerberus is not just a faster reactivity engine, but a significantly more efficient data layer than the industry standard.

Here is the exact breakdown of how our architecture dismantled TanStack Query across the board.

Our performance gains here are not accidental; they are the direct result of stripping away heavily abstracted observer patterns in favor of raw signal bindings and direct memory references.

Thanks!

This is another great release introducing some beneficial tools and overall improvements.

A special thanks to everyone who has helped validate the APIs, docs, and submitted features or bugfixes for this release.

There is no "I" in Cerber-"US"

Upgrading

To upgrade to this release, you can install the latest version of Cerberus React:

npm run up:cerberus

pnpm run up:cerberus

deno run npm:up:cerberus

bun run up:cerberus