Our take
- Activity
- 52
- Community
- 67
- Issues
- 48
- Pull requests
- 67
+167 stars in the last 30 days
- Slower to close issues than 99% of the projects we track
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
21,580Forks
3,058Change since Nov 15, 2026: +7 stars·+3 forks
Overview
FinGPT is a collection of tools and models for applying large language models to financial data. It addresses the cost and data-access challenges of building finance-specific models by emphasizing curated datasets, parameter-efficient fine-tuning, and reusable task benchmarks rather than training a foundation model from scratch.
The project is aimed at researchers and developers who want to evaluate or adapt models for financial NLP and related applications. Its scope includes sentiment analysis, information extraction, retrieval-augmented analysis, and a stock-movement forecasting demo. Models and datasets are also published through the FinGPT Hugging Face page.
Key Features
- Fine-tuning workflows using LoRA, including examples for sentiment models and other financial tasks.
- Financial instruction datasets for sentiment analysis, relation extraction, headline analysis, named-entity recognition, and question answering.
- Benchmarking resources for comparing open language models on financial datasets.
- Published task-specific and multi-task model adapters for several base-model families.
- FinGPT-Forecaster demo and dataset for analysis of Dow 30 stocks.
- FinGPT-RAG framework for retrieval-augmented financial sentiment analysis.
- Inference options that include local models and cloud LLM providers documented in the repository.
Use Cases
- Financial NLP researchers can reproduce or extend fine-tuning and benchmark experiments across financial datasets.
- Developers can adapt a model for classifying news or social-media sentiment when they have suitable data and compute.
- Teams building financial text-processing pipelines can explore relation extraction, headline classification, or entity recognition.
- Users exploring robo-advisory prototypes can try the Forecaster demo, keeping in mind its stated US Dow 30 market scope.
Getting Started
git clone https://github.com/AI4Finance-Foundation/FinGPT.git
cd FinGPT
pip install -r requirements.txt
pip install -e .
See the README and its SETUP guide for configuration, hardware guidance, and usage examples. A hosted Forecaster demo is available without local installation.
Alternatives
- FinRL: FinRL focuses on reinforcement-learning trading agents and market-data workflows, rather than finance-focused language models and NLP tasks.
- FinRL-Trading: FinRL-Trading builds stock-selection and trading strategies with supervised and reinforcement learning, rather than adapting language models for finance.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| FinGPT | Jupyter Notebook | MIT | 21.3k | Not checked yet |
| FinRL | Jupyter Notebook | MIT | 16.5k | Active |
| FinRL-Trading | Python | Apache-2.0 | 3.8k | Active |
Considerations
- Local inference and fine-tuning can require a suitable GPU; the README also describes cloud API alternatives.
- The repository covers multiple research and application components, so setup and requirements may differ by subproject. Check the relevant documentation before choosing a workflow.
- The Forecaster documentation describes support for US Dow 30 data. Applying it to other markets requires appropriate data and adaptation.
- Financial model outputs are not financial advice or trading recommendations. Validate results and consult qualified professionals before making investment decisions.
Found this useful?
Share it with someone who would like FinGPT.