Modelpublic

roadofriot/MindSparQ-Coder-1.5B

sourceHugging Faceapache-2.0updated 1mo agoView on Hugging Face

0likes252downloads

Model Card

🔮 MindSparQ-Coder-1.5B (2026 Frontier Vibe-Coding & Agentic Edition)

MindSparQ-Coder-1.5B is a production-ready, ultra-fast coding model fine-tuned for modern Vibe Coding, Software Architecture, and Autonomous Agentic Workflows (2026 Ecosystem).

Built by MindSparQ AI, this model combines fine-tuned specialized weights with lightweight footprint, enabling rapid inference on local commodity CPUs/GPUs with zero telemetry or code leakage.

🌟 Key Highlights

—⚡ Lightweight & Blazing Fast: Fits in ~1 GB RAM with 4-bit quantization (Q4KM) delivering ~15–30 tokens/sec on Intel Core i3 / Ryzen CPUs.
—🎨 Elite Vibe-Coding & Frontend Aesthetics: Trained on modern UI patterns (React Glassmorphism, Tailwind CSS, Dark Mode gradients, fluid animations).
—🛡️ Anti-Yes-Man Architectural Evaluation: Challenges insecure architectures (e.g. plaintext secrets, vulnerable sync loops) and proposes production-grade alternatives.
—🧭 Agentic Tool Calling & Planning: Structured to operate within multi-agent orchestration loops (Planner, Coder, Debugger, Reviewer).
—🔒 100% Local & Private: Run locally via llama.cpp, Ollama, or Python with zero outbound data leakage.

📦 Repository Files

File	Size	Description
`model.safetensors`	~2.9 GB	Full unquantized merged model weights (BF16/FP16)
`gguf/quantum_coder_q4_k_m.gguf`	~934 MB	Ultra-fast 4-bit quantized model for llama.cpp / Ollama
`gguf/quantum_coder_f16.gguf`	~2.9 GB	Full-precision 16-bit GGUF model
`tokenizer.json` / `config.json`	-	Standard Hugging Face Transformers configuration files

🚀 Quickstart Usage

1. Using LLaMA.cpp (Fast Local CPU Inference)

bash

./llama-cli -m gguf/quantum_coder_q4_k_m.gguf -p "<|im_start|>user\nWrite a FastAPI rate limiter in Python.<|im_end|>\n<|im_start|>assistant\n" -n 256 --threads 4

2. Using Python Transformers

python

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "roadofriot/MindSparQ-Coder-1.5B"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto")

prompt = "<|im_start|>user\nBuild a modern Glassmorphic CSS card token.<|im_end|>\n<|im_start|>assistant\n"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=256, temperature=0.3)
print(tokenizer.decode(outputs[0], skip_special_tokens=False))

📜 License

Apache-2.0 License. Powered by Qwen2.5-Coder architecture & MindSparQ AI Fine-Tuning.