We have theoretically recalibrated the extragalactic H II region sequence and defined the theoretical boundary between objects excited by stars, and those exited by active galactic nuclei. We used both the PEGASE and STARBURST99 codes to generate the SED of the young star clusters, and MAPPINGS 3 to compute photoionization models which include a self-consistent treatment of dust physics and chemical depletion. We find that the two stellar spectral synthesis codes are in excellent agreement with each other for zero-age instantaneous age models. Also, the extragalactic H II region sequence is reproduced remarkably well by our models, provided that the clusters which excite them are all rather young (<2 Myr). This is most likely an observational selection effect. The line ratio usually used for measuring the ionization parameter, [O III] λ5007/[O II] λλ3726, 3729, is indeed a good diagnostic. Amongst other easily observed line ratios we found that the [N II] λλ6548, 6584/[O II] λλ3726, 3729 ratio gives the best diagnostic of abundance, as it is monotonic between 0.1 and over 3.0 times solar metallicity. The extragalactic H II regions are characterized by abundances 0.1-2.5 solar and ionization parameters in the range 10⁷ < q < 10⁸ (or U ≡ q/c in the range -3.5 < log U < -2.5).