---
base_model: Qwen/Qwen2.5-0.5B-Instruct
tags: [gguf, ollama, llama.cpp, tax, taxes, accounting, tax-research, cpu, c0mpute]
license: apache-2.0
pipeline_tag: text-generation
---

# tax-guy-agent (GGUF)

> **Not a CPA. Not tax advice. Tax research from public sources; rules change every year and vary by place, so check with a CPA or enrolled agent before filing.**

[profullstack/tax-guy-agent](https://huggingface.co/profullstack/tax-guy-agent) merged into Qwen2.5-0.5B-Instruct and quantized (531 MB). Runs on a CPU.

```bash
ollama run hf.co/profullstack/tax-guy-agent-GGUF "Do I have to pay quarterly estimated taxes as a freelancer?"
```

Ollama picks up the system prompt (`system`) and sampling settings (`params`) from this repo.
llama.cpp: `llama-cli -hf profullstack/tax-guy-agent-GGUF`.

Trained on the c0mpute network: 2614 documents from rssamplifier.com topics and pit.world search, 400 LoRA steps (r=8), held-out perplexity 33.6 to 17.8. Adapter sha256 `821a038ac1ea43f804b795e795a308cfbebb141c24e3675223c29d80fb8db5b7` (matches the c0mpute receipt).
Every Infernet and c0mpute node pulls this model automatically ([house models](https://infernetprotocol.com/docs/house-models)).
It answers best with fresh search results in the prompt (`w3bs ask`, or the example's `chat.py`); a 0.5B model supplies the voice, not the facts.

Built with the [c0mpute example](https://github.com/profullstack/c0mpute/tree/master/examples/business-strategist-agent).