Microsoft

Phi 4 Mini Reasoning

Zero-eval
#1AIME

by Microsoft

+
+
+
+
About

Phi-4 Mini Reasoning was developed to incorporate extended thinking capabilities into the ultra-compact Phi-4 Mini architecture. Built to demonstrate that reasoning enhancements can be applied even to very small models, it brings analytical depth to resource-constrained environments.

+
+
+
+
Timeline
AnnouncedApr 30, 2025
ReleasedApr 30, 2025
Knowledge CutoffFeb 1, 2025
+
+
+
+
Specifications
Training Tokens150.0B
+
+
+
+
License & Family
License
MIT
Performance Overview
Performance metrics and category breakdown

Overall Performance

3 benchmarks
Average Score
68.0%
Best Score
94.6%
High Performers (80%+)
1
+
+
+
+
All Benchmark Results for Phi 4 Mini Reasoning
Complete list of benchmark scores with detailed information
MATH-500
text
0.95
94.6%
Self-reported
AIME
text
0.57
57.5%
Self-reported
GPQA
text
0.52
52.0%
Self-reported
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+