--10M dataset
with predictions as
(
select id ,probability
from table
)

-- give a rom_number to each record and then create 1000 groups
, bin_groups as (
select
id,
ceiling(1000.0*ROW_NUMBER() over(order by probability desc) / (select count(distinct id) from predictions)) as bins
from predictions
)

select *
from bin_groups
where bins = 1
limit 100

Resources exceeded during query execution: The query could not be executed in the allotted memory. Peak usage: 102% of limit. Top memory consumer(s): JOIN operations: 96% other/unattributed: 4%