Supermaven builds an AI coding assistant powered by an in-house generative model called Babble that supports a 1 million-token context window. The company says Babble uses a new neural architecture and in-house model serving infrastructure to deliver lower latency and it can process a developer's code repository in 10–20 seconds to learn APIs and conventions. Supermaven does not use customer data to train its models, retains customer data for a week to keep the system responsive, and says Babble was trained almost exclusively on publicly available code. Since launching in February, more than 35,000 developers have used Supermaven, the user base has grown 3x, and annual recurring revenue reached $1 million. The company currently has a five-person team and plans to spend funding on hiring developers and developing its text editor, which is in beta. Founder Jacob Jackson — who previously co-founded Tabnine and worked at OpenAI — says the growth and recent funding position Supermaven to scale through the end of the year.

Total raised

$12M

Funding rounds

1

Latest round

Equity

Latest activity

Sep 2024