diff --git a/Trained models/v1.1/README.md b/Trained models/v1.1/README.md
new file mode 100644
index 0000000..4ff57f7
--- /dev/null
+++ b/Trained models/v1.1/README.md	
@@ -0,0 +1,2 @@
+Trained with 30500 instances (~5 hours of gameplay)
+The dataset used is a merge between the dataset used for the v1.0 models and another dataset where the data has been recorded in a more blanced way, also we tried to drive better to record this new dataset. If we train the model with only the second dataset the results are prety bad, but the merge between the 2 datasets seems to improve the results.