Fireworks.ai, an AI company adapting its LL.M. to business needs

Fireworks.ai is a California-based artificial intelligence (AI) startup that provides unique solutions to enterprises. Rather than building a large language model (LLM) or base model from scratch, the AI ​​company will fine-tune the open source model and turn it into an application programming interface (API) to help enterprises deploy artificial intelligence in a seamless way Smart features. Fine-tuning narrows the scope of an AI model and focuses it on specific features. This allowed them to reduce the occurrence of AI hallucinations and significantly improve the model’s capabilities.

The artificial intelligence company was co-founded by Lin Qiao, who also serves as the company’s CEO.According to LinkedIn, Jo and her team founded the startup in October 2022 after serving as Meta’s senior director of engineering and working with artificial intelligence frameworks and platforms contour.in a dialogue She explained Fireworks.ai’s business model at TechCrunch, highlighting the fine-tuning services they offer. “It can be an off-the-shelf open source model, or it can be a model that we tune, or it can be a model that our customers can tune themselves,” she said. “All three varieties can be served through our inference engine API.”

This puts the company in a unique position where, while it is not innovating at the base model level, it is bridging the gap between LLMs and business-ready products that can be deployed seamlessly. Fireworks.ai primarily focuses on building APIs that allow its enterprise customers to plug and play any open source AI model in its catalog. The company is also letting businesses try out different AI models to choose one that suits their needs, according to the report.

at present, start up Claims to include 89 open source LLMs, such as Mixtral MoE 8x7B Instruct, Meta’s Llama 2 70B Chat, Google’s Gemma 7B Instruct, Stability AI’s Stable Diffusion XL, etc. The artificial intelligence company offers models either in a serverless format, which does not require enterprises to configure hardware or deploy the model, or as an on-demand model available for dedicated deployments, which is served on reserved GPU configurations based on business needs.

For on-demand formats, Fireworks.ai offers three payment plans – Developer, Business and Enterprise – with the Developer plan featuring a pay-per-use structure and a rate limit of 600 requests per minute, and the Enterprise tier with custom pricing offers and unlimited Interest rate limits. The serverless format is billed on a per-token pricing plan, where different models (depending on whether they are text-only, image-only, or multi-modal) will receive different prices.