An intercomparison of nine chemical mechanisms (e.g. ADOM, CBM-IV, EMEP, RADM2) as used by 12 contributing groups was conducted. The results for three scenarios are presented covering remote situations with a net O 3 loss of around 2.7 ppb (LAND and FREE) and a moderately polluted situation with O 3 formation of around 100 ppb (PLUMEI) over a 5 day simulation period. The overall tendencies (i.e. the total net production/loss over 5 days) for O 3 show a r.m.s. error of 38, 15 and 16%; for H 2O 2 the errors are 76, 23 and 30% (for LAND, FREE, PLUMEI). In terms of ozone production in PLUME1, the most productive mechanisms are EMEP and IVL, the RADM-type mechanisms lie in the mid-range and the CBM-1V type mechanisms fall at the bottom of the range. The differences in H 2O 2 can partly be explained by an incorrect use of the HO 2 + HO 2 rate constant and by differences in the treatment of the peroxy radical interactions. In the PLUME1 case the r.m.s. error of the PAN tendency was found to be 29%. Differences between mechanisms for the HO radical are 10, 15 and 19% and for the NO 3 radical 35, 16 and 40% (for LAND, FREE, PLUME1) in terms of the r.m.s. error of the results for a 12 h time period centred around the last noon (HO), respectively, a 8 h time period centred around the last midnight (NO 3) of simulation. Especially for NO 3 some differences are due to different numerical treatment of photolytic processes in the models. Large differences between mechanisms are observed for higher organic peroxides and higher aldehydes with a r.m.s. error of around 50% for the final concentration in PLUME1. The protocol of the intercomparison is given in the appendix, so that the comparison could be repeated for the purpose of mechanism development and sensitivity studies.