This contribution reports results for EE2 Test 1.12a and Test 1.12b which were proposed in JVET-AB0119. Test 1.12a proposes a 7-tap gradient and location based convolutional cross-component model (GL-CCCM) and Test 1.12b proposes a 5-tap gradient based CCCM. The proposed methods map luma values into chroma values when the prediction mode is activated by a PU level flag. The filter input in Test 1.12a consists of one spatial luma sample, two gradient values, two location information, a nonlinear term, and a bias term. The filter inputs for Test 1.12b are the same as Test 1.12a but without the location information. Filter coefficients are derived for each chroma block separately using regression based MSE minimization (i.e., the same solver as CCCM) on reference samples in the PU’s neighborhood. The impacts on coding efficiency and runtimes over ECM-7.0 are reportedly {for Y, U, V, EncT, DecT}: