Views
No views yet
NAMAN_HF_TOKEN.| 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. |
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.