Codestral 22B
by Mistral AI · mistral family
22B
parameters
text-generation code-generation
Codestral 22B is Mistral AI's dedicated code generation model, trained on over 80 programming languages. It excels at code completion, generation, and explanation tasks, with particular strength in Python, JavaScript, TypeScript, Java, C++, and many other languages. This model is optimized for developer workflows including fill-in-the-middle completion, test generation, and code documentation. It offers significantly better coding performance than general-purpose models of similar size.
Quick Start with Ollama
ollama run 22b-v0.1-q4_K_M | Creator | Mistral AI |
| Parameters | 22B |
| Architecture | transformer-decoder |
| Context Length | 32K tokens |
| License | Mistral AI Non-Production License |
| Released | May 29, 2024 |
| Ollama | codestral |
Quantization Options
| Format | File Size | VRAM Required | Quality | Ollama Tag |
|---|---|---|---|---|
| Q4_K_M recommended | 11.2 GB | 14.7 GB |
★
★
★
★
★
| 22b-v0.1-q4_K_M |
| Q5_K_M | 13 GB | 16.9 GB |
★
★
★
★
★
| 22b-v0.1-q5_K_M |
| Q8_0 | 19.8 GB | 24 GB |
★
★
★
★
★
| 22b-v0.1-q8_0 |
Compatible Hardware for Q4_K_M
Showing compatibility for the recommended quantization (Q4_K_M, 14.7 GB VRAM).
Compatible Hardware
5 hardware
device(s) cannot run this model configuration.
Benchmark Scores
60.0
mmlu