The model weights have also been made available along with the code.
A team from UC Berkeley, CMU, Stanford, and UC San Diego developed Vicuna-13B, an open-source replacement for GPT-4, after Stanford University released the ChatGPT clone Alpaca for $600. Vicuna-13B reportedly achieves 90% of ChatGPT’s quality, and the model’s training cost was around $300. LLaMA has been used to improve the model, and it now includes user-shared talks acquired from ShareGPT.
Early evaluations using GPT-4 as a judge have shown that Vicuna-13B surpasses other models like LLaMA and Stanford Alpaca in more than 90%* of cases and achieves over 90%* of the quality of both OpenAI’s “ChatGPT” and Google’s “Bard.” Additionally, Vicuna-13B’s performance is comparable positively to Stanford Alpaca and other open-source models. The findings have sparked a great deal of interest in the field of natural language processing, especially among companies looking to take advantage of the most recent developments in AI. The researchers wrote in a blog post that developing an evaluation system for chatbots is still an open problem that needs more investigation.
It will be interesting to see how it compares to other models like ChatGPT given the researchers’ audacious promises about its natural language processing abilities.
Vicuna-13B’s developers claim that it is capable of processing natural language better than other models like ChatGPT. While there are some similarities between the two versions, the model stands out due to its efficiency and customization options. Vicuna-13B’s performance is being eagerly watched by industry experts, who believe it will establish new standards for AI-powered language processing.
The model has some restrictions despite its outstanding powers. It has difficulty reasoning or performing mathematical calculations, for instance, and its outputs may not always be factually accurate.
The model still has to be fully improved to ensure safety or reduce any potential toxicity or bias. The developers have integrated OpenAI’s moderation API to filter out unsuitable user inputs in their online demo in order to allay these worries.
