NPY is NumPy's on-disk format for a single ndarray: a magic prefix, a header dict describing dtype/shape/order, and the raw element bytes. NPZ is just a ZIP that contains multiple .npy entries.

Can I download the values?

Yes. The JSON button exports the dtype, shape, and the first ~4,000 values to a JSON file you can paste back into Python with json.load + np.asarray.

Why are some int64 values rounded?

JavaScript's number type can only represent integers exactly up to 2^53. We surface 64-bit ints as numbers so they sort and render; if you need full precision use NumPy directly.

No. Parsing happens entirely in your browser using npyjs (and fflate for NPZ archives). Nothing leaves your device.

Does it work on mobile?

Yes. Tables scroll horizontally with a sticky row index; 3D+ arrays use a numeric stepper that's thumb-friendly.

Scientific.npy · .npz

NumPy NPY / NPZ viewer

ready

NumPy NPY / NPZ · .npy · .npz

Open NumPy NPY / NPZ files

NumPy array dumps.

Choose a fileor paste · ⌘V

About NumPy NPY / NPZ

NumPy array dumps.

NPY and NPZ are NumPy's native dump formats. .npy stores a single n-dimensional array with a tiny header describing dtype, shape, and memory order; .npz is a ZIP archive of one or more .npy files (what np.savez writes). This viewer surfaces the dtype, shape, and Fortran/C ordering, then renders the data: 1D arrays as a value list, 2D arrays as a true table with sticky headers, 3D+ arrays as a flat preview that you can step through one leading-axis slice at a time. NPZ bundles expose a tab per contained array. Big-integer dtypes (int64/uint64) are read as JS numbers — values beyond 2^53 will lose precision.

FAQ

What is an NPY file?: NPY is NumPy's on-disk format for a single ndarray: a magic prefix, a header dict describing dtype/shape/order, and the raw element bytes. NPZ is just a ZIP that contains multiple .npy entries.
Can I download the values?: Yes. The JSON button exports the dtype, shape, and the first ~4,000 values to a JSON file you can paste back into Python with json.load + np.asarray.
Why are some int64 values rounded?: JavaScript's number type can only represent integers exactly up to 2^53. We surface 64-bit ints as numbers so they sort and render; if you need full precision use NumPy directly.
Is my file uploaded?: No. Parsing happens entirely in your browser using npyjs (and fflate for NPZ archives). Nothing leaves your device.
Does it work on mobile?: Yes. Tables scroll horizontally with a sticky row index; 3D+ arrays use a numeric stepper that's thumb-friendly.

More Scientific