The American startup Goodfire, founded in 2024, has reached a valuation of $1.25 billion after raising $150 million in a Series B funding round. The investment was led by B Capital with participation from Menlo Ventures and Lightspeed Venture Partners. The company specializes in researching and interpreting the workings of artificial intelligence (AI) models.
This is reported by Finway
Goodfire and Other Market Leaders in AI Research
The market for startups focused on AI development and research is rapidly growing. Goodfire has joined a group of so-called neolabs that attract significant investments to study and develop interpretability of models. Other notable players include Thinking Machines Labs, founded by Mira Murati, and Safe Superintelligence, led by Ilia Sutskever, both of which have also raised billions of dollars. The new project Humans& has secured $480 million in funding, while AI researcher Richard Zoher is negotiating to raise funds for his lab, which is valued at $4 billion.
Interpretation of AI Models and Collaboration with Leading Companies
Goodfire’s CEO Eric Ho emphasizes the urgency of the rapid development of AI. The company is focused on developing interpretative models—tools that allow for the analysis and “mapping” of the structure of AI operations. This helps uncover new insights, refine models, and enhance their performance.
“I think what we are doing right now is quite reckless,” he said. “How can we trust and rely on what we do not understand?”
Developing interpretative models is a challenging task because the weights of models are usually encoded in data that is incomprehensible to humans. As a result, developers often have to retrain models completely without being able to effectively debug existing code. Goodfire is working to overcome this barrier by creating tools that help explain the mechanisms of artificial intelligence.
In practice, Goodfire’s technology has already proven effective: in the Prima Mente lab, an AI model was developed to predict Alzheimer’s disease, but the team did not understand how these predictions were formed. With the help of Goodfire’s interpreters, a new class of disease biomarkers was identified based on connections that the model established independently of human intervention.
Among Goodfire’s clients are Microsoft, Mayo Clinic, and the nonprofit Arc Institute. Some organizations use the lab to create new AI tools based on existing models, while others aim to enhance already developed products.
As of February 2026, the company has raised $209 million in investments. Eric Ho notes that Goodfire has already made significant progress in creating interpretative models for analyzing and debugging AI code. The funds raised will be directed towards improving model retraining methods, expanding computational capabilities, and increasing staff.
“I don’t like the current trajectory of AI development—we’re going to implement systems we don’t understand everywhere,” Ho said. “I think that’s bad, and I want to change it.”