Guide · Open source
How to use FinGPT: install, run the Forecaster demo and tips
Updated · written with AI from the project's own documentation and checked against our catalogue data
FinGPT is an open-source project from the AI4Finance Foundation that releases large language models trained or fine-tuned on financial text such as news and tweets. This guide explains the two easiest ways to use it: trying the FinGPT-Forecaster demo in a browser, and cloning the repository to work with the code locally. By the end you will know what the project is for, what you need to run it, and where its limits are.
FinGPTAI4Finance-Foundation/FinGPT · ★ 21.5k · MIT Install options & downloads Quick start
- 1Decide your path. Choose the browser demo if you only want to see Forecaster, or the local install if you want to work with the code.
- 2Open the Forecaster demo. Use the Hugging Face Space for FinGPT-Forecaster. No installation is needed.
- 3Enter a US ticker. Type a ticker such as AAPL, MSFT or NVDA, along with a yyyy-mm-dd start day.
- 4Set the news window and financials. Choose how many past weeks of news to retrieve and whether to include the latest basic financials.
- 5Clone the repository. Run git clone on the FinGPT repository and move into the FinGPT folder.
- 6Install dependencies. Run pip install -r requirements.txt, then pip install -e . to install the package in editable mode.
- 7Read SETUP.md. Check the project's SETUP.md for detailed local setup, hardware requirements and examples.
What FinGPT is and who it is for
FinGPT publishes its models on Hugging Face. The project also includes FinGPT-Forecaster, a 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 such as BloombergGPT.
- 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
Try the FinGPT-Forecaster demo without installing anything
The fastest way to see FinGPT work is the Hugging Face Space for FinGPT-Forecaster. No installation is required. Enter the inputs listed below and submit the form.
The README describes four inputs: a ticker symbol (for example AAPL, MSFT or NVDA), the day from which you want the prediction to happen in yyyy-mm-dd format, the number of past weeks of market news to retrieve, and whether to add the latest basic financials. After you click Submit, the demo returns an analysis of the company and a prediction for next week's stock price movement.
- Ticker symbol, e.g. AAPL, MSFT, NVDA
- Start day in yyyy-mm-dd format
- Number of past weeks of market news to retrieve
- Whether to include the latest basic financials
Install FinGPT locally by cloning the repository
To work with the code, clone the repository and install its dependencies. The README lists Python and pip as prerequisites. Run these commands from a terminal, in the folder where you want the project to live.
The README's Installation section installs the package in editable mode after the dependencies. Local models require a GPU, so skip this step for model work if you do not have one.
- Prerequisites: Python and pip
- GPU is required for running local models (per the README)
git clone https://github.com/AI4Finance-Foundation/FinGPT.git
cd FinGPT
pip install -r requirements.txt
pip install -e .Understand the basic usage options
The README's Basic Usage example shows two paths. The first uses a cloud API and does not need a GPU, by setting the environment variable FINGPT_LLM_PROVIDER to openai. The second loads local models with transformers and peft, which requires a GPU. The README points to SETUP.md for detailed examples, so treat the snippet as a starting outline rather than a complete program.
- Cloud API path: no GPU required, per the README example
- Local model path: requires a GPU, loads models with transformers and peft
- Detailed setup and hardware notes are in SETUP.md
import os
os.environ['FINGPT_LLM_PROVIDER'] = 'openai'Know the limits before you rely on the output
The README is clear that FinGPT-Forecaster is a research demo, not a trading tool, so its output is not investment advice. The current Forecaster is trained on US DOW 30 data. The README says other markets, including China, need their own data collection and fine-tuning.
The README's Why FinGPT section cites BloombergGPT training costs and says FinGPT fine-tuning can cost less than $300 per run. Those are the project's own claims and are not independently verified here.
- Forecaster output is a research demo, not investment advice
- Forecaster is trained on US DOW 30 data only
- Other markets require collecting data and fine-tuning the model
- Local models need a GPU
What people use it for
- Testing finance-tuned LLMs on news sentiment analysis
- Comparing open-source models on financial datasets for research
- Running the FinGPT-Forecaster demo to see how it structures a company analysis
- Studying how LoRA fine-tuning is applied to financial text
Frequently asked questions
Is FinGPT free to use?
The repository is licensed under the MIT licence, as shown on its GitHub page. Check the licence file in the repository for the full terms before you use or redistribute the code.
Does FinGPT require a GPU?
The README says local models require a GPU. The cloud API path shown in the Basic Usage example does not need one, and the Hugging Face Space runs the Forecaster demo without any installation.
How do I install FinGPT?
Clone the repository with git clone, enter the FinGPT folder, run pip install -r requirements.txt, then run pip install -e . The README lists Python and pip as prerequisites.
Can FinGPT predict stock prices for any market?
No. The README says FinGPT-Forecaster is trained on US DOW 30 data. Other markets, including China, need data collection and fine-tuning on that market.
Is FinGPT good for investment advice?
The README presents Forecaster as a research demo, not a trading tool. Do not use its output as investment advice.
How does FinGPT compare to BloombergGPT?
The README positions FinGPT as a more accessible, open-source alternative to BloombergGPT, which it describes as having privileged data access and APIs. It also claims much lower fine-tuning cost. Those are the project's claims, so compare them against your own tests.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
FinGPT suits developers, researchers and students who want open finance-tuned LLMs or a no-install look at the Forecaster demo. It is not a fit for investment decisions, non-US stock predictions, or anyone without a GPU who wants to run local models.
Install options & downloads