finance-cre-finetuned

Fine-tuned Phi-3.5-mini model for Commercial Real Estate (CRE) & Finance/Economics domain knowledge, built upon Microsoft Phi 3.5 mini.

Model Details

Attribute	Value
Base Model	microsoft/Phi-3.5-mini-instruct
Parameters	3.8B
Quantization	Q4_K_M (~2.2 GB)
Context Length	2048 tokens

Model Card

This model specializes in Commercial Real Estate terminology and concepts across all major asset classes.

Capabilities

• Financial Metrics: NOI, Cap Rate, DSCR, IRR, Equity Multiple, GAV, NAV, FFO, AFFO, Debt Yield
• Fund Structures: GP/LP, Preferred Return, Waterfall distributions, Carried Interest, Catch-up
• Debt Structures: Senior Loan, Mezzanine, Preferred Equity, A/B Notes, Cash Sweep, LTV
• Investment Strategies: Core, Core+, Value-Add, Opportunistic, Distressed
• Lease Terms: DTM, NTV, WALE, TI, CAM, Escalation, Base Year, Stop
• Asset Classes: Office, Industrial, Retail, Multifamily, SFR, BTR, Hospitality, DataCenter

Training Data

Trained on 658 Q&A pairs covering:

• CRE terminology definitions
• Financial calculations and formulas
• Investment analysis concepts
• Debt and equity structures
• Transaction types and strategies

Usage

Ollama (Recommended)

# Run directly from HuggingFace (no download needed)
ollama run hf.co/pramod-thaz/finance-cre-finetuned

# Or pull locally first
ollama pull pramod-thaz/finance-cre-finetuned
ollama run finance-cre-finetuned

llama.cpp

# Download GGUF and run with llama.cpp
./llama-cli -m finance-cre-finetuned-v0.1-Q4_K_M.gguf -n 256 --temp 0.3

Python (transformers)

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "pramod-thaz/finance-cre-finetuned"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

Version History

Version	Date	Notes
v0.1	April 2026	Initial release - 658 training pairs

Limitations

• Knowledge limited to training data (658 Q&A pairs)
• May not have real-time market data
• Designed for Q&A, not complex financial calculations
• Some specialized terms may require additional training

Disclaimer

This model is fine-tuned for educational and informational purposes. For actual investment decisions, consult with qualified professionals.

Citation

Fine-tuned Phi-3.5-mini for Commercial Real Estate
https://huggingface.co/pramod-thaz/finance-cre-finetuned

License

This model is based on Phi-3.5-mini which uses the MIT License. Check individual model cards for specific licensing information.

---

Note: This is v0.1 - a foundational model. Future versions will include more training data and improved accuracy on CRE-specific terminology.