Overview

Dask 2025.1, released on January 15, 2025, improves scheduler performance and integration with the PyData ecosystem.

Main Features

Optimized scheduler

The distributed scheduler has been optimized to reduce latency and better handle complex task graphs.

python 
import dask.dataframe as dd

# Parallel Parquet file reading
df = dd.read_parquet('data/*.parquet')
result = df.groupby('category').amount.mean()
print(result.compute())

Better PyData integration

Dask integrates better with NumPy 2.x and pandas 2.x for more seamless API compatibility.

python 
import dask.array as da
import numpy as np

# Distributed computation compatible with NumPy 2.x
x = da.random.random((10000, 10000), chunks=(1000, 1000))
result = x.mean(axis=0).compute()
print(f'Shape: {result.shape}')

Sources