---
license: apache-2.0
tags:
- mamba
- codestral-mamba
- qwen2.5
- state-space-model
- benchmark
- evaluation
---

# MAMBA_7B: Codestral Mamba 7B Installation, Benchmarks & Architectural Research

This repository contains the complete codebase, benchmark suite, evaluation datasets, research documents, and visual PowerPoint presentations for **MAMBA_7B**.

Published to **namanadep** Hugging Face profile using `NAMAN_HF_TOKEN`.

---

## 📊 Summary Benchmark Metrics (Codestral Mamba 7B vs. Qwen 2.5 7B)

| Metric | Codestral Mamba 7B | Qwen 2.5 7B Instruct | Takeaway |
| :--- | :---: | :---: | :--- |
| **Architecture** | **Selective State Space Model (SSM S6)** | **Multi-Head Self-Attention Transformer** | Mamba eliminates $O(N^2)$ quadratic KV-cache memory scaling. |
| **Average Latency** | **4.28s** | **7.40s** | **42.2% faster completion** for Codestral Mamba. |
| **Generation Speed** | **194.8 t/s** | **193.3 t/s** | Identical throughput on NVIDIA H200 GPUs. |
| **Memory Footprint** | **Constant $O(1)$ Memory State** | $O(N)$ Growth | Fixed VRAM up to 256k long-context reasoning. |

---

## 📂 Repository Layout & Uploaded Artifacts

- `docs/CODESTRAL_MAMBA_7B_VS_QWEN_7B_COMPARISON.md`: 71 KB Exhaustive 10-Prompt Benchmark Report.
- `docs/CODESTRAL_MAMBA_7B_VS_QWEN_7B_COMPARISON.pptx`: 9-Slide Visual Benchmark Comparison Deck.
- `docs/MAMBA_7B_INSTALLATION_AND_ARCHITECTURE_GUIDE.pptx`: 8-Slide Hands-on Installation & Architecture Journey Deck.
- `docs/MAMBA_MODELS_RESEARCH_OLLAMA_HUGGINGFACE.md`: State Space Models Architectural Research Document.
- `data/mamba_vs_qwen_results.json`: Raw Evaluation JSON transcripts across 10 technical categories.
- `src/`: Complete Python benchmark test harness and slide generation scripts.