Fireworks.ai: An open source API Democratizing Generative AI for developers

Nearly everyone is vying for a slice of the generative AI market. While the spotlight often shines on major players like OpenAI, Anthropic, Cohere, and tech giants such as Microsoft, Meta, Google, and Amazon, many startups are also tackling the generative AI challenge.

One such startup is Fireworks.ai. Although it may not have the same level of brand recognition as some of its competitors, it boasts the largest open-source model API, with over 12,000 users according to the company. This significant open-source adoption tends to attract investor interest, with the company having raised $25 million to date.

Lin Qiao, co-founder and CEO of Fireworks, emphasizes that her company is not focused on training foundational models from scratch. Instead, it specializes in fine-tuning existing models to meet the specific needs of businesses. "It can be either off-the-shelf, open-source models or the models we tune, or the models our customers can tune by themselves. All three varieties can be served through our inference engine API," Qiao told TechCrunch.

As an API, developers can integrate Fireworks into their applications, use their preferred model trained on their data, and quickly add generative AI capabilities, such as asking questions. Qiao highlights that the platform is fast, efficient, and produces high-quality results.

Another advantage of Fireworks' approach is its support for experimenting with multiple models, a crucial capability in a rapidly evolving market. "Our philosophy here is we want to empower users to iterate and experiment with multiple models and have effective tools to infuse their data into multiple models and test with a product," she explained.

Even more crucially, they control costs by limiting the model size to between 7 billion and 13 billion parameters, a stark contrast to ChatGPT4's over 1 trillion parameters. This limitation reduces the scope of words the large language model can comprehend but allows developers to concentrate on smaller, specialized datasets tailored for more specific business applications.

Qiao's background uniquely positions her to develop such a system, having previously led the AI platform development team at Meta. Her goal was to construct a rapid, scalable development engine to drive AI across all of Meta's products and services. Drawing from this experience, she has created an API-based tool that grants such power to any company without necessitating the extensive engineering resources of a company like Meta.

In 2022, Fireworks secured a $25 million funding round led by Benchmark, with support from Sequoia Capital and angel investors such as Databricks and Snowflake. These strategic investments from Databricks and Snowflake are noteworthy, as both companies specialize in data storage tools, and Fireworks' platform enables users to leverage their data effectively.