NEA Leads Funding of Startup That Helps Customers Cut AI Costs; A Chinese Startup Seizes on Meta’s Open-Source Llama Model

In September, I wrote about an emerging group of startups that help developers manage the exploding costs of AI. These companies figure out what tasks can be handled by cheap-to-use AI models so their customers don’t end up paying for a more expensive model from OpenAI that they don’t really need. 

There’s signs these companies have legs, at least in the eyes of venture capitalists. One of the earliest of these startups, San Francisco-based Martian, has just raised $9 million in seed funding co-led by NEA and Prosus Ventures to fund its mission. Carya Venture Partners and General Catalyst also participated in the round.

The five-person “model routing” startup was founded by Shriyash Upadhyay and Etan Ginsberg, two University of Pennsylvania graduates who researched AI at the university. They’ve developed a model router that automatically sends requests to different models depending on a customer’s needs, including cost and speed.