Mistral AI

Mistral Small 3 24B Base

Multimodal
Zero-eval
#1AGIEval

by Mistral AI

+
+
+
+
About

Mistral Small 24B Base was developed as a 24-billion-parameter foundation model, designed to serve as a base for fine-tuning and customization. Built to provide a strong starting point for domain-specific applications, it represents an intermediate-scale option in Mistral's model lineup.

+
+
+
+
Timeline
AnnouncedJan 30, 2025
ReleasedJan 30, 2025
Knowledge CutoffOct 1, 2023
+
+
+
+
Specifications
Capabilities
Multimodal
+
+
+
+
License & Family
License
Apache 2.0
Performance Overview
Performance metrics and category breakdown

Overall Performance

9 benchmarks
Average Score
67.0%
Best Score
91.3%
High Performers (80%+)
4
+
+
+
+
All Benchmark Results for Mistral Small 3 24B Base
Complete list of benchmark scores with detailed information
ARC-C
text
0.91
91.3%
Self-reported
GSM8k
text
0.81
80.7%
Self-reported
MMLU
text
0.81
80.7%
Self-reported
TriviaQA
text
0.80
80.3%
Self-reported
MBPP
text
0.70
69.6%
Self-reported
AGIEval
text
0.66
65.8%
Self-reported
MMLU-Pro
text
0.54
54.4%
Self-reported
MATH
text
0.46
46.0%
Self-reported
GPQA
text
0.34
34.4%
Self-reported
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+