Model weights: Qwen2.5-Coder-1.5B-Instruct (Apache 2.0) — a strong 1.5B coding model.
Agent engine: the Nexus Coder framework (nexus/) — 60+ skills and 80+ tools with automatic registration, agent planner/router/memory/safety, data pipeline, and training utilities.

🚀 Quick Usage

The model is a standard Qwen2ForCausalLM. Load it with transformers:

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "AdminReal/NexusCoder"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

messages = [{"role": "system", "content": "You are Nexus Coder, a helpful coding assistant."},
            {"role": "user", "content": "Write a Python function to compute fibonacci."}]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer([text], return_tensors="pt")
out = model.generate(**inputs, max_new_tokens=512)
print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))

CLI chat demo (from source):

python scripts/chat.py --model AdminReal/NexusCoder

📁 Repository Contents

Part	What it is	License
`model.safetensors`, `config.json`, `tokenizer.*`	Qwen2.5-Coder-1.5B-Instruct weights	Apache 2.0
`nexus/`	Agent engine source (skills, tools, agent, data, optim)	NAL-1.0
`configs/`	Experimental architecture designs (`tiny` → `423b`) — not the hosted model	NAL-1.0
`docs/`, `scripts/`, `tests/`	Documentation, CLI scripts, tests	NAL-1.0

The configs/nexus_coder_*.yaml files describe a from-scratch MoE research architecture and are independent from the Qwen2-based weights hosted in this repo.

⚖️ Licenses

Model weights: Apache 2.0 (from Qwen/Qwen2.5-Coder-1.5B-Instruct). See LICENSE.
Source code (nexus engine): NexusCoder Attribution License v1.0 (NAL-1.0). Attribution required to Hieu Louis (github.com/mhieuhonda).

👤 Author

Hieu Louis · 2026

GitHub: @mhieuhonda
Project: NexusCoder