In this paper I review some recent advances in the use of large amounts of atomic data in the modelling of atmospheres and winds of hot stars. The review is highly selective but representative of current developments. A more general overview is to be found in Kudritzki and Hummer (1990) although the field is changing so rapidly that much has happened since then. The paper breaks down into three parts: work on line formation, in which the atmospheric structure is known and held fixed, is described first, then follows a description of the inclusion of line opacities in non-LTE in the atmosphere problem itself, and finally recent developments in the theory of radiatively driven stellar winds are summarized. Here special emphasis is given to a novel distance determination method based entirely on spectroscopie quantities. I close with a brief shopping list.

In a series of papers, Becker and Butler (1992,1994a, b,c) have investigated iron and nickel spectra in sub-dwarfs using the complete linearization method of Auer and Heasley (1976). The method scales linearly with the number of frequency points so they were able to use well over 10000 frequencies to adequately describe the line opacities. Several thousand lines were treated explicitly and the resultant computed spectra gave execellent fits to observed Hubble spectra in the wavelength ranges dominated by the ions concerned.

The different ionization stages gave consistent results for the iron and nickel abundances but only after line-blocking from millions of spectral lines in the far UV had been included. This was done using the Kurucz (1988) line lists coupled with line grouping as suggested by Anderson (1989) and described briefly in the next section.

The line-blanketed atmospheres of Kurucz (1991) are the best available up to about 30000K, where non-LTE effects start to become important. Non-LTE line-blanketed atmospheres have become feasible because the computational requirements of the accelerated lambda iteration (ALI) method (Werner and Husfeld, 1985) also scale linearly with the number of frequency points. On the other hand, Anderson (1989) suggested grouping energetically adjacent atomic levels together to form pseudo-levels on the basis that although they might, as a group, be in non-LTE, they should be in LTE with respect to one another due to the large number of collisions between them. This greatly reduces the number of levels to be considered but instead gives rise to highly complicated pseudo line-profiles. Grigsby et al (1992), who did not use ALI, constructed the first grid of line-blanketed non-LTE models by using a variation on the Opacity Distribution Function concept to group line opacities into blocks thereby reducing the number of frequency points required. Dreizler and Werner (1993) on the other hand were able to sample the opacity as they used ALI in their models.