Yes, matlabs default tolerances all now seem to be abstol = 1e-10, reltol = 1e-6. This matches what octave uses for quadgk, with the exception that for 'single' types octave uses 1e-5 and 1e-4. Matlab keeps the higher tol for singles and the online documentation just warns that you may need to use a larger value for convergence. The thought was that octave would apply its current behavior with all the other functions, making sure it's detailed in all the docs, maybe highlighted with a 'known incompatibility' note.