20+ Open Source ChatGPT Alternatives for Commercial and Research Use: A Comprehensive Database

Hrant Davtyan
3 min readMay 5, 2023

--

Image 1. Screenshot of the AirTable database with 20+ ChatGPT alternatives

Are you excited to build a new product or a feature using generative AI? Want to explore alternative open-source AI language models for your commercial or research needs? Look no further! We have compiled a comprehensive database of 20+ ChatGPT alternatives that are available for you to fine-tune and utilize. Read on to discover the best options for your unique needs.

With the increasing popularity of ChatGPT, more and more AI products are being built on top of its API. However, the limited customization options for ChatGPT due to its closed-source nature has led to the emergence of open-source alternatives such as LLaMA or Dolly. In this article, we have compiled a comprehensive database of 20+ open-source ChatGPT alternatives that can be customized to meet your specific commercial or research needs.

Image 2. “ChatGPT Alternatives” sign held by a Lama, generated by DeepFloyd IF

Insights from the database

We are releasing the database on AirTable so that you can interactively filter and choose the models that meet your needs. While putting together the database, we observe several key insights that are summarized below.

  1. EleutherAI models, including GPT NeoX, GPT J, and Pythia, are among the most popular open-source models that can also be used for commercial purposes. Some of those models can be customized at an affordable cost (less than $300), and some may even fit a single GPU machine.
  2. If you need a large open-source model, BLOOMZ by Hugging Face (et al.) is based on the original BLOOM model and has 176 billion parameters. Although it requires more resources to train and fine-tune, it may be worth considering if you require a more powerful model.
  3. Apart from English, Chinese is the second most researched language among the open-source models in our database. Raven is the only fully RNN-based model in the list, while all others follow the transformer architecture with an attention mechanism. Most of the models follow the decoder-only architecture, with exceptions such as mT0 or FastChatT5.
  4. Google and Anthropic are the only companies in our database that offer closed-source ChatGPT alternatives, while Raven is the only model that is using RNNs (fully attention-free mechanism).

The full interactive database can be found here.

Closing thoughts

We will continue to update our database with more information, including training cost, training data, and performance evaluation. Additionally, we will be releasing similar databases for foundational language and image models in the future, so stay tuned!

If you want to add a new model to our database or report any errors, please contact us at info@anania.ai or refer to our GitHub repo.

By exploring these open-source ChatGPT alternatives, you can fine-tune language models to meet your specific needs, whether you’re working on a commercial project or conducting research (be careful in your choice though!). From LLaMA to BLOOMZ, there’s a wide range of models available, and we hope our database helps you identify the best options for your use case.

--

--

Hrant Davtyan

Founder @ Anania. I am writing about Data analytics, Generative AI, LLMs, Machine Learning.