Entering a matrix 01 Set Rows and Cols (1 to 6); the grid resizes and keeps the entries that still fit. 02 Type one number per cell. Minus signs and scientific notation are accepted; a blank or non-numeric cell is highlighted red and blocks the run. 03 Pick the operation — Det, Inverse, Transpose, Rank, RREF, Eigen, LU, QR, Pinv, eᴬ, A×B or A+B — then press Compute . The two-sided operations also reveal matrix B. 04 Read the result. Scalars print as a value, matrices print one row per line, and decompositions label their factors (L:, U:, Q:, R:). 05 Reuse the answer. A ⇐ loads a matrix from the workspace, and the result's ports store or send the value onward.
Verified readouts In 11 2 / 3 5 the slash separates rows; every readout below reproduces here.
Operation
Input
Readout
Det
11 2 / 3 5
49
Inverse
the same A
0.1020 -0.0408 / -0.0612 0.2245 , printed to four decimals
Rank
1e-11 0 / 0 1e-11
2 with (zero tolerance 4.44e-27, scales with the matrix magnitude)
Eigen
2 1 / 1 2
Eigenvalues: 1, 3
eᴬ
0 1 / 0 0
1 1 / 0 1
The determinant is exact (11·5 − 2·3), and the inverse is (5, −2; −3, 11) divided by 49, so 0.1020 and 0.2245 are those fractions rounded to four decimals. The exponential is exact too: 0 1 / 0 0 squares to zero, so eᴬ = I + A.
Where the numbers stop
Singular is called singular. 1 2 / 2 4 returns determinant 0 with the warning Matrix is singular or numerically near-singular: the determinant may be pure rounding noise — treat it as 0; Inverse answers Matrix is singular — the inverse does not exist.
Display rounds; the stored value does not. On-screen matrices print non-integers to four decimals — the inverse above shows 0.1020 — while anything you copy, store or send keeps full double precision (5/49 = 0.10204081…).
Ill-conditioned inverses are flagged. A large condition number earns a note about lost significant digits instead of an exact-looking inverse; a determinant can be honest noise long before it prints as 0. What it does not do: it shows no elimination or factor steps, and eigenvalues are numeric, not symbolic. Use the equations panel for a linear system, Step-by-step for the rules, the geometry panel for coordinate work.
Rank uses a scaled zero tolerance The signature trap is the rank readout: it prints its own cutoff, (zero tolerance 4.44e-27, scales with the matrix magnitude), instead of using a fixed threshold. A diagonal matrix of 1e-11 entries is still rank 2 (tolerance 4.44e-27 ), while entries around 1e8 with a dependent row read rank 1 (tolerance 1.78e-7 ). A small pivot is not automatically zero, and a large-looking residue is not automatically a new pivot — read the tolerance before interpreting the number.
Where it fits Solving a linear system with RREF Enter the augmented matrix 1 1 1 6 / 0 2 5 -4 / 2 5 -1 27 (3 × 4) and choose RREF: the identity block reads 1 0 0 5 , 0 1 0 3 , 0 0 1 -2 , the solution (5, 3, −2).
Checking an eigen readout 2 1 / 1 2 gives Eigenvalues: 1, 3 ; the 90° rotation 0 -1 / 1 0 gives Eigenvalues: 0+1i, 0−1i . Complex pairs print in a+bi form, and a repeated eigenvalue is returned the expected number of times.
Privacy Every operation runs in your browser; the entries, factors and results are never uploaded.
References
Wikipedia, Rank (linear algebra) , en.wikipedia.org (访问日期:2026-10-01)— rank, pivots and numerical thresholds.
Wikipedia, Eigenvalues and eigenvectors , en.wikipedia.org (访问日期:2026-10-01)— real and complex eigenvalues.
Wikipedia, Matrix exponential , en.wikipedia.org (访问日期:2026-10-01)— the series and the nilpotent case.
Wikipedia, Condition number , en.wikipedia.org (访问日期:2026-10-01)— why an inverse warning depends on conditioning.
Calculators in this hub
Hand-picked tools, one click away. The mini versions compute live and carry your values into the full calculator.
Sources & review
Reviewed by CalcX Editorial Team
Updated 2026-10-01