Cosine Similarity Calculator

Measuring Direction, Not Magnitude

Cosine similarity ignores how long two vectors are and measures only the angle between them — two documents with very different word counts can still be judged highly similar if the proportions of their vocabulary point in the same direction. That magnitude-independence is exactly why it's the standard similarity measure for text embeddings and other high-dimensional vector comparisons.

The Formula

cos(θ) = (A · B) / (|A| × |B|)

The result ranges from −1 (pointing in exactly opposite directions) through 0 (orthogonal, no relationship) to 1 (pointing in exactly the same direction). This calculator also converts the result into the actual angle in degrees.

Where This Matters

  • Document and text similarity — comparing TF-IDF or embedding vectors to find related documents regardless of length.
  • Recommendation systems — measuring similarity between user preference vectors or item feature vectors.
  • Clustering and nearest-neighbor search — cosine similarity is a common distance metric in high-dimensional spaces where Euclidean distance behaves poorly.

Worked Example

Cosine similarity for A = [1, 2, 3], B = [4, 5, 6]
QuantityValue
A · B (dot product)32
|A| (magnitude)3.742
|B| (magnitude)8.775
cos(θ)0.974632
Angle θ≈ 12.93°

cos(θ) = 32 / (3.742 × 8.775) = 0.9746; the small angle confirms the vectors point in nearly the same direction.

How to Use This Calculator

  1. Enter Vector A as a comma-separated list of components.
  2. Enter Vector B with the same number of components.
  3. Select Calculate to get the cosine similarity and the angle between the vectors.

Related Calculations

For a distance-based rather than angle-based comparison, use the Euclidean Distance Calculator, or standardize vectors first with the Feature Scaling Calculator.