Hugging Face's logo

0sz1
/
IT-auditor-1.7B

GGUF
English
legal
contract-analysis
smollm2
unsloth
privacy
conversational
Model card Files
xet
 Community
1
How to use from
llama.cpp
Install (macOS, Linux) 
curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf 0sz1/IT-auditor-1.7B:Q8_0
# Run inference directly in the terminal:
llama cli -hf 0sz1/IT-auditor-1.7B:Q8_0
Install from WinGet (Windows) 
winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf 0sz1/IT-auditor-1.7B:Q8_0
# Run inference directly in the terminal:
llama cli -hf 0sz1/IT-auditor-1.7B:Q8_0
Use pre-built binary 
# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf 0sz1/IT-auditor-1.7B:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf 0sz1/IT-auditor-1.7B:Q8_0
Build from source code 
git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf 0sz1/IT-auditor-1.7B:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf 0sz1/IT-auditor-1.7B:Q8_0
Use Docker 
docker model run hf.co/0sz1/IT-auditor-1.7B:Q8_0
Quick Links

IT-auditor-1.7B-Alpha

IT-auditor-1.7B-Alpha is a specialized, privacy-first Small Language Model (SLM) designed to audit IT contracts and find "lethal" clauses for software developers and freelancers.

It is built on top of SmolLM2-1.7B-Instruct and fine-tuned on a high-density dataset of real-world tech agreements.

Key Features

  • Privacy-First: Designed to run 100% offline on consumer hardware (GTX 1650/3050). No data leaves your machine.
  • SEC EDGAR Grounding: Trained on verified filings from top tech companies (Apple, NVIDIA, Palantir).
  • Expert Distillation: Insights distilled from Llama-3-70B experts specifically for IT-law risk detection.
  • High Performance: Optimized for 4096 context window and blazing fast inference (60+ tokens/sec).

What it detects:

  • Uncapped Liability: Finds clauses where you risk your entire business.
  • Aggressive IP Transfers: Detects when you lose rights to your ideas before payment.
  • Unilateral Termination: Points out power imbalances in notice periods.
  • Payment Traps: Flags ambiguous "sole discretion" payment terms.

Benchmark (Alpha Preview)

In internal tests, IT-auditor detected 24-hour termination traps and "unpaid work" IP transfers that general-purpose 7B models often gloss over.

image

image

image

image

image

Usage (GGUF)

Recommended for use with LM Studio, Ollama, or llama.cpp.

System Prompt:

"You are a legal auditor. Your task is to find traps in contracts that hurt the developer and point them out clearly."

Research Background

This model is part of a broader study on Information Asymmetry and Semantic Entropy in recursive AI systems. For technical details, see my papers on Zenodo.

Contact & Custom Solutions

Need a custom model tuned on your proprietary corporate data? Looking for a 8B/70B high-precision version?

Disclaimer: This model is a research tool and does not constitute legal advice. Always consult with a human lawyer for final contract decisions.

Downloads last month
14
GGUF
Model size
2B params
Architecture
llama
Hardware compatibility
Log In to add your hardware

8-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for 0sz1/IT-auditor-1.7B

Base model

HuggingFaceTB/SmolLM2-1.7B
Quantized
(54)
this model