connectedtechco/syrup-code-x0.1-4b

Syrup Code x0.1 4b

Syrup Code x0.1 4b is an experimental coding model fine-tuned from Qwen3 4B Instruct with Connected Tech codebase structure, specialized for generating structured code in Go, TypeScript, React, and Svelte.

Quantized GGUF variants: syrup-code-x0.1-4b-f16, syrup-code-x0.1-4b-q4_k, syrup-code-x0.1-4b-q8_0

Usage

Command Line (llama.cpp)

hf download connectedtechco/syrup-code-x0.1-4b gguf/syrup-code-x0.1-4b-(VARIANT).gguf --local-dir .
./llama-server --host 0.0.0.0 --port 8080 -m gguf/syrup-code-x0.1-4b-(VARIANT).gguf

Python (Transformers)

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("connectedtechco/syrup-code-x0.1-4b")
tokenizer = AutoTokenizer.from_pretrained("connectedtechco/syrup-code-x0.1-4b")

prompt = "write a svelte component for navigation menu"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_length=512)
print(tokenizer.decode(outputs[0]))

Training Details

• Training Data: Connected Tech codebase
• Training Method: LoRA fine-tuning

Limitations

• Experimental model, may produce incorrect output
• Optimized for Connected Tech codebase structure, intended for internal use only.

Citation

@misc{syrup-code-x0.1-4b,
  author = {Bhumjate S.},
  title = {Syrup Code x0.1 4b},
  year = {2025},
  publisher = {HuggingFace},
  url = {https://huggingface.co/connectedtechco/syrup-code-x0.1-4b}
}