San Francisco, February 5, 2026 – Goodfire, the San Francisco-based AI research lab dedicated to decoding the “black box” of neural networks, has officially reached unicorn status. The startup announced today it has raised $150 million in a Series B funding round, catapulting its valuation to $1.25 billion just months after its Series A.
As AI models become increasingly integrated into society, the inability to understand their internal reasoningoften referred to as the “black box” problem—has become a major barrier to safety and trust. Goodfire specializes in mechanistic interpretability, a field of science focused on reverse-engineering neural networks to understand how specific neurons and “features” drive model behavior.
The Series B round was led by B Capital, with a star-studded list of participants including Salesforce Ventures, DFJ Growth, and former Google CEO Eric Schmidt. Existing backers Menlo Ventures, Lightspeed Venture Partners, and Anthropic also doubled down on their support. This latest infusion brings Goodfire’s total capital raised to $209 million.
Real-World Impact: From Hallucinations to Health
Goodfire’s flagship platform, Ember, is already proving that understanding a model’s “thoughts” leads to better performance.
- Hallucination Reduction: Goodfire reported a 50% reduction in hallucinations in a large language model by using “feature steering”—a process where specific internal mechanisms are directly edited to keep the model on track.
- Scientific Discovery: In a groundbreaking collaboration with Prima Mente and the Mayo Clinic, Goodfire used its tools to identify a novel class of Alzheimer’s biomarkers from an epigenetic foundation model. This marks one of the first major scientific discoveries derived from reverse-engineering an AI’s internal logic.
Unlike traditional AI companies that focus on “more data and more compute,” Goodfire is building a “Model Design Environment.” This allows engineers to treat AI models like traditional software—where they can inspect, debug, and modify code with precision.
“I think what we’re doing right now [in the industry] is quite reckless,” said Eric Ho, co-founder and CEO of Goodfire, in an interview with Bloomberg. “How can we trust and rely on something that we don’t understand? We are building a microscope for the digital mind.”
The new capital will be used to scale research into AI agents and life sciences, while also expanding its team of researchers—many of whom joined from OpenAI and Google DeepMind.
Key Highlights:
- Unicorn Status: Goodfire raised $150 million at a $1.25 billion valuation, officially joining the ranks of AI unicorns in early 2026.
- Major Backing: The round was led by B Capital, with participation from Salesforce Ventures and industry veteran Eric Schmidt.
- Ember Platform: Its core tool, Ember, allows developers to “steer” AI behavior, reducing hallucinations by 50% through direct internal edits.
- Medical Breakthrough: Goodfire’s tech recently helped identify new Alzheimer’s biomarkers, proving that AI interpretability can accelerate natural sciences.