If you need this often, and/or the count is an integral part of your Tab1 model, you should use a hybrid property such as described in the other answer. If on the other hand you need this just for a single query, then you could just create the scalar subquery using Query.label(), or Query.as_scalar():

count_stmt = session.query(func.count(1)).\
    filter(Tab2.tab1_id == Tab1.id).\
    group_by(Tab2.col1).\
    label('cnt')

session.query(Tab1, count_stmt).filter(...).limit(100)

The subquery will automatically correlate what it can from the enclosing query.

You can do this, but it works in a quite different way to how you have written it. You can create a property of Tab1 which depends on the relationship to tab2 (assuming that tab2.tab1_id is a foreign key, which it should be.

Your models look like this:

class Parent(Base):
    __tablename__ = 'parent'
    id = Column(Integer, primary_key=True)
    children = relationship("Child")

class Child(Base):
    __tablename__ = 'child'
    id = Column(Integer, primary_key=True)
    parent_id = Column(Integer, ForeignKey('parent.id'))

as per the docs on relationships

Then you can add something like

@hybrid_property
def number_of_children(self):
    if self.children:
        return len(self.children)
    return 0

@number_of_children.expression
def number_of_children(cls):
    return (select([func.count(Child.id)])
            .where(Child.cover_id == cls.id))

to the Parent model, as per this answer and more docs.

Once you've done this, you can filter on this property the same as any other column-based one.