Context : Solar twins are objects of great interest in that they allow us to understand better how stellar evolution and structure are affected by variations of the stellar mass , age and chemical composition in the vicinity of the commonly accepted solar values .

Aims : We aim to use the existing spectrophotometric , interferometric and asteroseismic data for the solar twin 18 Sco to constrain stellar evolution models .
18 Sco is the brightest solar twin and is a good benchmark for the study of solar twins .
The goal is to obtain realistic estimates of its physical characteristics ( mass , age , initial chemical composition , mixing-length parameter ) and realistic associated uncertainties using stellar models .

Methods : We set up a Bayesian model that relates the statistical properties of the data to the probability density of the stellar parameters .
Special care is given to the modelling of the likelihood for the seismic data , using Gaussian mixture models .
The probability densities of the stellar parameters are approximated numerically using an adaptive MCMC algorithm .
From these approximate distributions we proceeded to a statistical analysis .
We also performed the same exercise using local optimisation .

Results : The precision on the mass is approximately 6 % .
The precision reached on X _ { 0 } and Z _ { 0 } and the mixing-length parameter are respectively 6 % , 9 % , and 35 % .
The posterior density for the age is bimodal , with modes at 4.67 Gyr and 6.95 Gyr , the first one being slightly more likely .
We show that this bimodality is directly related to the structure of the seismic data .
When asteroseismic data or interferometric data are excluded , we find significant losses of precision for the mass and the initial hydrogen-mass fraction .
Our final estimates of the uncertainties from the Bayesian analysis are significantly larger than values inferred from local optimization .
This also holds true for several estimates of the age encountered in the literature .

Conclusions :