FinGPT
Open-source large language models tuned for financial tasks like sentiment analysis and stock forecasting.
- Stars
- 21.5k
- Forks
- 3k
- Licence
- MIT
- Language
- Jupyter Notebook
- Last update
- 12 days ago
- Type
- AI model / agent
What it is
FinGPT is a project that releases open-source large language models (LLMs) trained or fine-tuned on financial text, such as news and tweets. Its models are published on Hugging Face, and the project also includes a FinGPT-Forecaster demo that produces an analysis and a next-week stock movement prediction for a ticker you enter. The README says the goal is a more accessible alternative to closed, costly finance models like BloombergGPT.
- Run financial sentiment analysis on news or tweets using a released FinGPT model
- Try FinGPT-Forecaster in the Hugging Face Space by entering a ticker and date
- Fine-tune a base LLM on new financial data using the LoRA approach described in the README
- Explore the FinGPT-Benchmark results to compare models on financial datasets
Great for
- Developers who want to experiment with finance-tuned LLMs for sentiment analysis
- Researchers comparing open-source models on financial datasets
- People who want to try the FinGPT-Forecaster demo without installing anything
- Students learning how LoRA fine-tuning is applied to finance data
Not ideal for
- Anyone looking for investment advice: the README presents Forecaster as a research demo, not a trading tool
- Users needing non-US stock predictions: the README says Forecaster is trained on US DOW 30 data and other markets need their own fine-tuning
- People without a GPU who want to run local models: the README says local models require a GPU
How to get it
- 1
Clone and build from source
You'll need: Python · pip · GPU for running local models (per README)
Clone the repository
git clone https://github.com/AI4Finance-Foundation/FinGPT.gitEnter the project folder
cd FinGPTInstall dependencies
pip install -r requirements.txtInstall the package in editable mode
pip install -e . - 2
Run it
- Clone the repo and install dependencies as shown in the Installation section
- Read SETUP.md for hardware requirements and local-run details
- Try the FinGPT-Forecaster demo on the Hugging Face Space if you do not want to install anything
- For local models, load them with transformers and peft as shown in the Basic Usage example and SETUP.md
- 3
Get the most out of it
- Use the Hugging Face model pages to pick a released FinGPT model for your task
- Read the FAQ.md for usage capabilities and limitations
- Check the FinGPT-Benchmark folder to compare models on financial datasets
- Review the FinGPT-Forecaster folder for customized implementations
Downloads
Strengths
- Models and datasets are publicly available on Hugging Face
- MIT license allows broad reuse
- Backed by published papers and an active README with a clear changelog
Trade-offs
- The README notes that much of the Wall Street data and APIs remain closed, so the project depends on public data
- Setup for local models requires a GPU and points to a separate SETUP guide for details
- The README is truncated and the benchmark table is cut off, so some performance claims cannot be verified here
How it compares
| Project | Stars | Type | Level | Updated | Best for |
|---|---|---|---|---|---|
| FinGPT (this) | 21.5k | AI model / agent | advanced | 12 days ago | Developers who want to experiment with finance-tuned LLMs for sentiment analysis |
| FinRL | 17k | Library | intermediate | 12 days ago | Students learning how reinforcement learning is applied to financial markets |
| annotated_deep_learning_paper_implementations | 67.6k | Learning resource | intermediate | 4 days ago | Students learning how transformers, GANs, or optimizers work internally |
| stable-baselines3 | 14k | Library | intermediate | 11 days ago | Researchers who want a reliable baseline to compare a new method against |
| DocsGPT | 18.4k | App | intermediate | 4 days ago | Teams that want an internal assistant over their own documents and wikis |
| wandb | 11.3k | Library | beginner | 4 days ago | Machine learning engineers who want to compare training runs over time |
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.