Swap two rows
Rᵢ ↔ RⱼMove a usable pivot into the next pivot position.
Enter integers, decimals, or scientific notation. Display precision does not round intermediate calculations.
Rows are numbered from 1. When enabled, the vertical divider marks the augmented final column.
Rᵢ ↔ RⱼMove a usable pivot into the next pivot position.
Rᵢ ← cRᵢ, c ≠ 0This calculator scales each pivot to 1 for a clearer echelon form.
Rᵢ ← Rᵢ + cRⱼEliminate entries below a pivot for REF and both below and above it for RREF.
10⁻⁴⁰ through 10⁴⁰ in magnitude. The selected relative tolerance controls numerical pivot decisions; use exact-arithmetic software when exact rational output is required.[A | b].The default matrix represents three equations in three unknowns:
[[1, 2, −1 | −4], [2, 3, −1 | −11], [−2, 0, −3 | 22]][[1, 0, 0 | −8], [0, 1, 0 | 1], [0, 0, 1 | −2]]The three coefficient columns are pivots, so the coefficient rank equals the number of variables. The system has the unique solution x = −8, y = 1, and z = −2.
A pivot column contains a leading entry in an echelon form. The number of pivots is the matrix rank. Non-pivot coefficient columns are free columns; in a consistent system, each free variable contributes a parameter to the solution family.
For an augmented matrix [A | b], the system is inconsistent when reduction produces a row of the form [0 0 … 0 | c] with nonzero c. Otherwise, it has a unique solution when every variable column is a pivot and infinitely many solutions when at least one variable column is free.
REF has zeros below each pivot and places successive pivots farther right. RREF additionally makes each pivot 1 and clears every other entry in its pivot column.
Yes. Every matrix has one unique RREF. REF is not unique, so another valid elimination order may give a different echelon matrix with the same rank and pivot structure.
Yes. This calculator accepts rectangular matrices from 1×1 through 8×9. A matrix does not need to be square to have REF, RREF, rank, pivots, or free columns.
The last column is treated as the constants vector rather than a variable column. It still participates in every row operation, but the calculator does not choose a coefficient pivot from it.
Yes. Each elementary row operation is reversible, so it replaces the system with an equivalent system having the same solutions.
Most decimal fractions are approximated in binary floating point. The selected tolerance suppresses numerically negligible pivot candidates, but poorly scaled or sensitive matrices may still require more specialized numerical software.
Yes. The page parses and reduces the matrix locally, and creates copied text and downloads in your browser. The calculator code does not transmit or save entered values.
Last reviewed: August 5, 2026 by the Starlight Tools editorial team.
The algorithm selects the strongest available pivot relative to each candidate row, scales pivots to 1, and records every effective elementary row operation. REF output depends on the pivot strategy; RREF is mathematically unique, subject to the selected numerical tolerance and floating-point rounding.