Follow Datanami:
February 21, 2024

Predibase Launches LoRA Land to Rival GPT-4


Predibase, the leading developer platform for fine-tuning LLMs, has introduced LoRA Land, a collection of 25 open-source fine-tuned models that the company claims can challenge or even outperform the hugely popular GPT- 4.0 by OpenAI. 

LoRA Land is powered by Predibase’s serverless fine-tuned endpoints and open-source LoRAX framework. The new platform offers a wide range of use cases from sentiment analysis to summarization. 

With GPT-4 being one of the most widely used LLMs around the globe, it is a daunting challenge for LoRA Land to outperform it. However, Predibase seems confident in the ability of its newest offering. The company claims that LoRA Land offers a significantly more cost-effective way for organizations to train highly accurate and specialized GenAI applications. 

As the cost of building GPT models or fine-tuning LLMS from scratch is astoundingly high, using specialized LLMs is becoming a popular alternative, and this is exactly where Predibase is likely to position LoRA Land in the competitive landscape. 

Using small and more specialized LLMS, developers leverage techniques such as parameter-efficient fine-tuning and low-rank adaptation to create high-performing AI applications to lower the costs of fine-tuning LLMs. Predibase says that it has incorporated such techniques in its platform, offering users the option of choosing the most appropriate LLM for their use case and fine-tuning it accordingly. 

One of the reasons why fine-tuned LLMS have historically been so expensive to put into production is that they require a dedicated GPU for each model. For users that need to deploy LLMS to address various use cases, the GPU expenses accumulate to become a major hurdle for growth and innovation. While the initial experimentation with LLMs accessed with APIs is relatively inexpensive, the expense rises quickly when a full-scale implementation is deployed. 

Fine-tuning open-source LLMs is not only expensive from a resource perspective but there is also the major issue of the lack of AI skills, which is reported by many as one of the primary hurdles to AI adoption. 

Predibase overcomes the cost challenge by designing LoRA Land to serve multiple fine-tuned LLMs on a single graphics processing unit. According to Predibase, the 25 LLMs in LoRA Land were fine-tuned at an average GPU cost of less than $8. Not only is this cheaper, but users also don’t have to wait for a cold GPU to start spinning before serving each model. Other advantages offered by LoRA Land include a highly scalable infrastructure model and instant deployment and prompting. 


“Organizations are increasingly recognizing the benefits of having many smaller, fine-tuned models for different use cases and customers,” said Dev Rishi, co-founder and CEO of Predibase. According to internal data, 65% of surveyed organizations plan on deploying two or more fine-tuned LLMs in the next 12 months, and 18% plan on deploying 6 or more.”

The efficiency and affordability offered by LoRA Land have leveled the playing field for smaller companies in the AI race. It is a game changer for companies that want to deploy a wide range of specialized LLMs to power their business. 

The introduction of the new platform not only offers technological innovation, it has the potential to transform the landscape of AI development. With its high-performing, cost-effective, and accessible, Predibase has set new standards for the industry. 

Related Items 

Privacy and Ethical Hurdles to LLM Adoption Grow

What Does ChatGPT for Your Enterprise Really Mean?

Key Insights from GenAI and ML Report