benchmarks

altor-vec benchmark comparisons

These pages compare altor-vec with hosted vector databases, embedded engines, and browser-capable alternatives. The comparisons are intentionally honest: altor-vec does not win on backend scale or infrastructure features, but it does win when local UX, privacy, and small-bundle delivery are the primary constraints.

Browse all 10 benchmark pages

Summary table

Benchmark methodology

These measurements reflect altor-vec running in a controlled browser environment. All queries execute against a pre-built HNSW index loaded from a JSON file — no embedding generation is included in the latency numbers. Embeddings are generated once at build time.

Test configuration

altor-vec latency results

What these numbers mean for your app

A p50 of 0.4ms and p95 of 0.8ms means search is effectively instant from the user's perspective — human perception of "instantaneous" begins around 100ms. At sub-millisecond latency, the bottleneck shifts to rendering results, not computing them.

For comparison, network-dependent search adds a baseline of 20–150ms for the round-trip, before the server executes its own query. A cloud search call at 100ms total latency is 125× slower than an altor-vec local query at p95. Whether that matters depends on your product: for autocomplete-as-you-type, the difference is significant; for triggered search (user presses Enter), it is less critical.

The 17MB memory footprint for 10K vectors at 384 dimensions fits comfortably within modern browser memory budgets. For 100K vectors at 384 dimensions, expect approximately 170MB — viable for desktop, worth testing on mobile.

Run your own benchmark

import init, { WasmSearchEngine } from 'altor-vec';
await init();

const vectors = new Float32Array(N * DIM); // your embeddings
const engine = WasmSearchEngine.from_vectors(vectors, DIM, 16, 200, 50);
const query = new Float32Array(DIM); // your query embedding
const iterations = 1000;
const times = [];

for (let i = 0; i < iterations; i++) {
  const start = performance.now();
  engine.search(query, 5);
  times.push(performance.now() - start);
}

times.sort((a, b) => a - b);
console.log('p50:', times[Math.floor(iterations * 0.5)].toFixed(2) + 'ms');
console.log('p95:', times[Math.floor(iterations * 0.95)].toFixed(2) + 'ms');
console.log('p99:', times[Math.floor(iterations * 0.99)].toFixed(2) + 'ms');

Run this in your browser against your own index to get accurate numbers for your hardware, vector dimensions, and index size.

When to use altor-vec vs a server-side solution