Improve model description: features, strengths, personal AI benefits

README.md

@@ -13,13 +13,21 @@ tags:
 
 # zindango-slm
 
-A generic instruction-following model for Zindango that **only responds in English** and never generates Chinese.
+A lightweight, capable instruction-following model for Zindango. Fine-tuned for clarity, versatility, and personal AI workloads.
 
 ## Features
 
-- **Identity**: Identifies as zindango-slm, the Zindango model
-- **Generic use**: Broad instruction-following for Zindango use cases
-- **No Chinese**: Never generates Chinese—rejects Chinese prompts and refuses requests to output Chinese
+- **Task-agnostic**: Handles summaries, Q&A, drafting, analysis, and open-ended assistance
+- **Consistent identity**: Reliably introduces itself as zindango-slm, the Zindango model
+- **English-optimized**: Tuned for natural, coherent responses in English
+
+## Why zindango-slm for Personal AI
+
+- **3B parameters** — Runs on consumer hardware (CPU, modest GPUs, edge devices) without cloud dependencies
+- **Compact and fast** — Low latency for real-time conversations and local inference
+- **Privacy-preserving** — Run entirely on-device; no data leaves your machine
+- **Customizable base** — Easy to further fine-tune for your own workflows and preferences
+- **GGUF support** — Use with llama.cpp for efficient CPU inference and broad compatibility
 
 ## GGUF (llama.cpp)