An experimental version of Stable LM 3B has been launched today by its creators. This latest addition to their suite of high-performance generative AI solutions has 3 billion parameters, making it compact enough to operate on portable digital devices. The makers of Stable LM 3B are excited about its capabilities and portability.
Like the previous Stable LM release, a key advantage of Stable LM 3B is its smaller size and efficiency compared to models with 7 to 70 billion parameters typically used in the industry. These smaller models require fewer resources and have lower operating costs, making them highly accessible for most users. Not only does this greater affordability have financial benefits, but it also makes the models more environmentally friendly since they consume much less power. Despite its compact size, Stable LM 3B is highly competitive, outperforming previous 3 billion parameter models and even some of the best 7 billion parameter open-source language models.
The development of Stable LM 3B expands the range of advanced conversational applications that can be deployed cost-effectively on the edge or home PCs. This enables individuals and companies to create technologies like creative writing assistance while maintaining low costs and high performance.
Compared to the earlier Stable LM release, this latest version significantly improves text generation while keeping fast execution speed. It demonstrates enhanced performance on common natural language processing benchmark tests, including those measuring common sense reasoning and general knowledge. To achieve these gains, Stable LM 3B has undergone extensive training across multiple epochs on high-quality data, surpassing its predecessors at similar model sizes.
While Stable LM 3B is a general language model, it can also be fine-tuned for specialized uses like programming assistance. This could allow companies to customize it cost-effectively for their own data, whether as a customer service chatbot, coding assistant, etc.
However, its creators caution that Stable LM 3B is a base model requiring safety adjustments before deployment in applications like chat interfaces. Developers must evaluate and fine-tune the model based on their specific use case. Safety testing of the instruction fine-tuned version is currently underway, with plans to release it soon.
The creators firmly believe smaller customizable models like Stable LM 3B will play an increasing role in practical generative AI applications. They encourage the AI community to try the model by downloading the weights from the Hugging Face platform. This initial release is under the open-source CC-By-SA 4.0 license.
https://huggingface.co/stabilityai/stablelm-3b-4e1t