Open-source LLM Tracker

Last updated: 24th April 2023

🚀 Open-source Large Language Models

Over recent months and weeks, the open-source community have been really active. They have been trying to make AI as accessible and as open as possible. There have been a flurry of datasets, base models and fine-tuned models released. The pace has been electric and exciting!

This tracker is a live document which attempts to keep track of the datasets, base models and fine-tuned models. This document is meant to help the legal community make sense of all the recent events and keep an eye on trends.

❓Why is this important?

In the legal space, open-source models can enable Legal AI applications using internal data that would otherwise be impossible because of privacy and security constraints. Keeping track of these events in the open-source community, will help us understand what options are available from a technical standpoint.

The tracker keeps track of:

1. 🥒 Datasets
2. 🍞 Base Models
3. 🥧 Fine-tuned Models

Alongside a description of each, we've also tried to track licensing information about the data and models so you can understand how the different datasets and models could be used and whether they could be used for commercial purposes.

🥒 Datasets

🍞 Base-Model Suites

🥧 Fine-tuned Models

🔮 Legal AI and Open-source

Here at simplexico, we believe that in order for Legal AI applications to be realised within law firms and legal teams, open-source models are the way forward. Despite much of the recent talk around AI, LLMs and Generative AI bringing change to the legal industry, there hasn't been much conversation around the how.

How do we practically train AI models on internal data? How do we leverage the best-in-class LLMs for our own use cases?

At simplexico, we are doing our bit to make sense of the how question when it comes to the design, implementation and delivery of bespoke Legal AI solutions. We offer a range of white-glove services around Legal AI education, design and development. If you are interested in educating yourselves about LLMs or are looking to fine-tune an open-source LLM on your own data we'd love to chat. You can find us at [email protected].