Overview
Scikit-learn 1.8, released on December 10, 2025, adds Array API GPU support, free-threaded mode and temperature calibration.
Main Features
Array API GPU and free-threaded
Array API compatible estimators can now run on GPU via CuPy or PyTorch. Free-threaded mode enables true parallelism.
python
from sklearn.linear_model import LogisticRegression
import numpy as np
X = np.random.randn(1000, 10)
y = (X[:, 0] > 0).astype(int)
clf = LogisticRegression()
clf.fit(X, y)
print(f'Score: {clf.score(X, y):.2f}')Temperature calibration
The new temperature calibration adjusts predicted probabilities for more reliable confidence scores.