Goodfire, a San Francisco AI interpretability lab, raised a $150m Series B round at a $1.25bn valuation, announced on 5 February 2026. B Capital led the round, according to SiliconANGLE and Venture Capital Access Online. The raise came less than a year after its Series A.
Existing investors Juniper Ventures, Menlo Ventures, Lightspeed Venture Partners, South Park Commons and Wing Venture Capital took part. New investors included DFJ Growth, Salesforce Ventures and Eric Schmidt, the former Google chief executive.
“Interpretability, for us, is the toolset for a new domain of science”Eric Ho, CEO, Goodfire
What Goodfire does
Goodfire researches how AI models represent and use concepts internally, and builds tools to inspect and change them. Its website describes methods to understand model mechanisms, catch unexpected behaviour from training runs and reduce hallucinations, plus work in life sciences on interpretable models.
SiliconANGLE described a planned platform that Goodfire calls a model design environment. Its first component covers training, showing how a model learns skills from its dataset and identifying flaws. The second monitors models in production. Goodfire says it cut hallucinations by half in one recent project, a result the sources do not independently verify. Last year the company developed a method called SPD, which removes model components one by one to see whether they affect output.
Healthcare AI startup Prima Mente is one of its first customers. SiliconANGLE said Goodfire's analysis found that Prima Mente's Alzheimer's detection model mainly considers the length of cfDNA fragments. Other partners named by Venture Capital Access Online are Mayo Clinic and Arc Institute.
Team and funding so far
Eric Ho is chief executive. Tom McGrath is a co-founder who founded the interpretability team at Google DeepMind, and the team includes Nick Cammarata, a core contributor to OpenAI's interpretability team, and Leon Bergen, a University of California San Diego professor on leave. Goodfire is a public benefit corporation.
The company says it is backed by more than $200m in total. The sources do not state the Series A amount, a founding year or a headcount.
What the money is for
Goodfire will use the funds for frontier research, the next generation of its core product, and partnerships in AI agents and life sciences, according to Venture Capital Access Online. SiliconANGLE said the money will enhance the platform and fund research projects.
Ho said interpretability is the toolset for a new domain of science, a way to form hypotheses. Yan-David Erlich, a general partner at B Capital and former chief operating officer at Weights & Biases, said Goodfire is unlocking the ability to truly steer what models learn and make them safer and more useful. In Venture Capital Access Online's account he added that the industry is building the most consequential technology of our time. Pulse 2.0 quoted Ho saying that every engineering discipline has been gated by fundamental science, and reported that the team includes researchers from DeepMind, OpenAI, Harvard and Stanford.
Company website: goodfire.com