This article will summarise the evaluation of four CFD software codes (CFX, FDS4, SMAFS and SOFIE). Evaluation was performed by means of comparing the simulation data with experimental scenarios. The scenarios were chosen to represent scenarios frequently assessed using firesafety design based on performance. The greatest difficulty during validation of CFD-codes is to find well documented relevant experiments. It is true that many large scale experiments including smoke spread have been published but many of them are not described sufficiently and are therefore impossible to use for validation. It is very unusual that data on reproducibility and repeatability is available. After an extensive discussion between the participants in the project five scenarios were chosen. Scenarios 1A (large room with vents), and scenarios 1B1 and 1B2 (corridor with vents) were well documented experiments (repeatability) with ceiling vents and ceiling jets performed earlier at SP in Borås. Scenario 2 is a tunnel fire taken from an experimental study in Boston, US. Scenario 3 is an atrium fire taken from an experimental study in Japan. Lack of well documented fires in retail premises lead us to perform a series of tests in 1/2 scale, 4.1 and 4.2 at SP in Borås. Scenario 5 is an experimentalstudy from Australia including smoke spread from a fire room to an adjacent corridor and a room. The conditions for the simulations included information about the scenarios given above. No information about experimental results was available until the results from the simulations had been sent in to the project manager. The intention was to imitate the working conditions for a consultant as far as possible when using a CFD-simulation to estimate fire safety. In this manner the simulations are dependent on many factors such as the CFD-code used, judgement of input data of the scenarios done by the operator, operator skill etc. The results from the evaluation of the CFD codes showed that the simulation results generally were a good description of the experimental fires studied, but that limitations and specific properties of the different codes, together with the operator factor, can strongly influence the results. The detailed results of all comparisons are given in the final report of the project. It was not possible to include these detailed evaluations in this conference paper but these results are available in the main report as well as in future scientific articles.