Codestral-22B
#1HumanEvalFIM-Average
#1Spider
#1HumanEval-Average
+2 more
by Mistral AI
+
+
+
+
About
Codestral 22B was developed as a specialized coding model from Mistral AI, designed to excel at code generation, completion, and understanding tasks. Built with 22 billion parameters optimized for programming, it serves developers requiring advanced assistance with software development across multiple programming languages.
+
+
+
+
Timeline
AnnouncedMay 29, 2024
ReleasedMay 29, 2024
+
+
+
+
License & Family
License
MNPL-0.1
Performance Overview
Performance metrics and category breakdown
Overall Performance
7 benchmarks
Average Score
65.9%
Best Score
91.6%
High Performers (80%+)
2+
+
+
+
All Benchmark Results for Codestral-22B
Complete list of benchmark scores with detailed information
| HumanEvalFIM-Average | text | 0.92 | 91.6% | Self-reported | |
| HumanEval | text | 0.81 | 81.1% | Self-reported | |
| MBPP | text | 0.78 | 78.2% | Self-reported | |
| Spider | text | 0.64 | 63.5% | Self-reported | |
| HumanEval-Average | text | 0.61 | 61.5% | Self-reported | |
| CruxEval-O | text | 0.51 | 51.3% | Self-reported | |
| RepoBench | text | 0.34 | 34.0% | Self-reported |
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+