Making NLP accessible to


We develop new and unique types of state-of-the-art large language models which are designed to have high performance at a low cost.


Our new and innovative technology allows you to reduce your large language model (LLM) compute costs no matter what, regardless of your current solutions.


We are just an API call away. We want to enable easy integration into existing NLP pipelines while making it easy to create new ones.


Despite being cheaper than everyone else, performance will not be sacrificed. Our accuracies will be comparable to the widely-used state-of-the-art models.

Our Mission

At Formic, we are building a new architecture for Large Language Models that will change the landscape of AI. Using our innovative technology, we want to create products that can help change the world, one byte at a time.

Our long-term mission is to one day create the neural side of Artificial General Intelligence (AGI). To complete our goal, we are scaling our technology around 3 pillars:

  • Cost: We want to enable new innovation through attainable costs
  • Integration: We want to ensure that anyone, not just data scientists, can use our technology
  • Performance: We build our models to extend their capabilities beyond standardized benchmarks

Frequently Asked Questions

  • How will your model differ from the “traditional” Transformer model?

    Our model is built on an entirely new architecture that is entirely different from popular Transformer models like GPT-3 and BERT. It will be cheaper and more efficient than anything out there on the market while still providing state-of-the-art performance. Our new architecture will allow us to change the field of language models and hopefully bring AGI one step closer.

  • We are currently looking to validate the scaling of our technology prior to releasing the system source to the public under open-source. Nevertheless, we are always happy to open up academic partnerships with institutions looking to move NLP and AGI forward.

  • We believe that bias inherently exists in all datasets, and therefore, providing visibility into what creates the outputs of a model will allow us to navigate the consequences. As opposed to some models in this space, we are looking to make our model truly explainable; i.e., we can provide descriptive analysis on why the model made a particular decision for all predictions.

  • From the integration standpoint, our function library is a simple API call away. In terms of use cases, we’ve seen the technology used in a number of creative ways from simple document analysis all the way to creating dynamically changing video games. Reach out to us for a demo and we’ll be more than happy to discuss how this technology will help improve your business.

  • We will work with your company to make sure that we can effectively either work into existing or create new NLP pipelines. Our team will sit down with yours to organize and plan the proper integration strategy to make sure that we uphold best practices for your policies and regulations.