StabilityAI’s StableLM: a promising alternative to OpenAI’s GPT-4?

A new open source linguistic modeling toolset

The large linguistic models sector continues to grow. StabilityAI, maker of the popular Stable Diffusion image generation tool, has released a suite of open source linguistic modeling tools.

What is StableLM?

Dubbed “StableLM,” the alpha versions of the publicly available package currently contain models with 3 billion and 7 billion parameters, with 15 billion, 30 billion and 65 billion parameter models “in development” and a 175 billion model planned for the future.

Does size matter?

In comparison, GPT-4 has an estimated parameter count of one billion, six times higher than its predecessor GPT-3.

However, parameter counts may not be a uniform measure of LLM effectiveness, as StabilityAI pointed out in the blog post announcing the launch of StableLM:

“StableLM is trained on a new experimental dataset built on The Pile, but three times larger with 1.5 billion content tokens […] The richness of this dataset gives StableLM surprisingly high performance on conversational and scheduling tasks, despite its small size of 3 to 7 billion parameters.”

How robust are the StableLM models?

It is unclear at this time to what extent the StableLM models are robust. The StabilityAI team noted on the organization’s Github page that more information on LM capabilities, including model specifications and training settings, would be forthcoming soon.

Interested in trying StableLM?

Provided the models work well enough in testing, the arrival of a powerful open source alternative to OpenAI’s ChatGPT could prove interesting for the cryptocurrency trading world.

As we told you earlier, people are building advanced trading bots using the GPT API and new variants that incorporate access to third-party tools, such as BabyAGI and AutoGPT.

Adding open source models to the mix could be a boon for tech-savvy traders who don’t want to pay OpenAI access premiums.

StableLM test on HuggingFace

Interested parties can test a real-time interface to the StableLM model of 7B parameters hosted on HuggingFace.

In summary, StabilityAI’s release of StableLM is an exciting development in the world of large open source language models. While it is not yet clear how robust these models are, the ability to have access to tools of this magnitude and sophistication may be a boon to tech-savvy traders looking to avoid the high costs of accessing third-party tools. In addition, StableLM’s focus on performance in conversational and scheduling tasks may have valuable applications in areas such as cryptocurrency trading. In short, we will have to closely follow the development of StableLM and see how it affects the landscape of large linguistic models.