RUN THIS LLM
Search local LLM hardware requirements
Codestral Mamba 7B
Mistral · 7B · Code
Code model using Mamba architecture instead of Transformer. Efficient for long code contexts.
VRAM Requirements
| Quantization | VRAM |
|---|---|
| Q4_K_M (smallest) | 4.2 GB |
| Q8_0 (balanced) | 7.7 GB |
| FP16 (full quality) | 14 GB |
Specifications
- Parameters: 7B
- Category: Code
- Max context: 256K tokens
- System RAM: 8 GB minimum
- HuggingFace: mistralai/Codestral-Mamba-7B-v0.1
Benchmarks
- HumanEval: 56 — Code generation (pass@1)
Loading interactive analysis...