I think this is not convergence problem. let me know the keywords that you are using.

opt=cartesian gen pseudo=cards rm062x scf=tight

I introduce the opt=cartesian to avoid problems with the structure.

I advise you to optimize your "big" structure initially at the b3lyp/lanl2dz level of theory [# b3lyp/lanl2dz opt=Cartesian]. Then you can reoptimize it using larger basis set (you can actually use Stuttgart basis set and pseudo potentials for Pt, which are available at the EMSL library: https://bse.pnl.gov/bse/portal). See below the example of .com (input) file that you should use in order to complete your calculations properly.

%chk=*specify direction of your .chk file from b3lyp/lanl2dz calculations*

%Mem=10GB

%NProcShared=2

#B3lYP/GEN guess=read geom=check pseudo=read opt scf=tight

Name

0 1

H 0 -(elements that your structure contains except Pt. In this example they are hydrogen (H) and oxygen (O))

6-311++G**

****

O 0

6-311++G**

****

Pt 0

S 3 1.00

16.5595630 -0.8849447

13.8924400 1.5011228

5.8536080 -1.5529012

S 1 1.00

1.2873200 1.0000000

S 1 1.00

0.6047320 1.0000000

S 1 1.00

0.1427830 1.0000000

S 1 1.00

0.0509690 1.0000000

S 1 1.00

0.0150000 1.0000000

P 2 1.00

7.9251750 4.9530757

7.3415380 -5.8982100

P 2 1.00

1.9125150 0.3047425

1.0715450 0.7164894

P 1 1.00

0.4379170 1.0000000

P 1 1.00

0.0936210 1.0000000

P 1 1.00

0.0278020 1.0000000

D 4 1.00

3.9395310 -0.5826439

3.5877770 0.5922576

1.2862310 0.4736921

0.5198140 0.5765202

D 1 1.00

0.1747150 1.0000000

D 1 1.00

0.0500000 1.0000000

****

PT 0

PT-ECP 5 60

h-ul potential

1

2 1.000000000 0.000000000

s-ul potential

2

2 13.428651000 579.223861000

2 6.714326000 29.669491000

p-ul potential

2

2 10.365944000 280.860774000

2 5.182972000 26.745382000

d-ul potential

2

2 7.600479000 120.396444000

2 3.800240000 15.810921000

f-ul potential

1

2 3.309569000 24.314376000

g-ul potential

1

2 5.277289000 -24.218675000