Chlorophyll a is often used as a proxy to estimate marine phytoplankton biomass given its optical properties that can be easily measured both in situ and remotely. The Chl:C ratio which is used to convert from chlorophyll to biomass, is, however, not constant in phytoplankton. It depends on temperature, light and nutrients, and is assumed to be regulated by the cells to maximise the growth rate under limiting environmental conditions. This process, called acclimation, increases the chlorophyll content under low light and decreases it under nutrient limitation. The Geider model that allows for dynamic Chl:C ratios is now included in an increasing number of marine ecosystem models. However, the ratios are seldomly validated and their effect on net primary production estimations from chlorophyll data is still highly uncertain. The aim of this study is to compare the Chl:C ratio in eight model simulations from the MARine Ecosystem Model Intercomparison Project (MAREMIP) and/or the Coupled Model Intercomparison Project Phase 5 (CMIP5), i.e. REcoM2, TOPAZ, MEM, NOBM, PlankTOM5.3, BEC, CNRM-PISCES and IPSL-PISCES. We focus on the annual climatology of the ratio in surface waters for the period 2000-2005. To assess the ability of these simulations to represent the chlorophyll field, we first compare model outputs with satellite observations. It appears that the main patterns in chlorophyll distribution are modelled quite well in the open ocean, but not in coastal areas, partly due to the coarse resolution of these model runs. The core analysis of this study consists in investigating the reasons behind the discrepancies in the Chl:C ratio among models. We highlight that models have different ways of dealing with this ratio: MEM uses a constant ratio, the ratio in NOBM is only light-dependent, and the remaining models allow for a variable ratio. Nevertheless, they all agree on the fact that the slope of the linear regression between phytoplankton carbon and chlorophyll is a number smaller than 1. In the group of models considering a variable ratio, the annual climatology of the Chl:C ratio simulated by REcoM2 substantially differs. We demonstrate that the Chl degradation rate is probably too high in this model run and that the minimum tolerated values of the ratio are too low. Eventually, an interesting feature about REcoM2 is that it can handle unbalanced growth conditions. Some of the other models are only valid under the hypothesis of balanced growth, which rarely happens in natural environments. We hope this present study brings some enlightenment on the variability of the phytoplankton Chl:C ratio and will contribute to improve our estimations of primary production and by this means, future climate projections.