Goodfire
Goodfire Inc. enhances its AI interpretability platform with $150M Series B funding.
About
Goodfire Inc. is a startup focused on uncovering how artificial intelligence models make decisions. They have developed a platform called a model design environment to map out LLMs' internal components, aiding in understanding data processing, identifying flaws, and improving output quality.
Similar companies
InstaDeep is a global AI leader specializing in decision-making AI systems for enterprises, with roots in Africa and offices worldwide.
In Decision Intelligence
AI startup developing brain-inspired AGI with ultra-low compute requirements.
In Frontier AI Lab
AI models that learn adaptively and use less computing power through gradient-free learning.
In Adaptive AI
Hugging Face is the leading AI community and platform for machine learning collaboration, enabling developers to build, share, and deploy models, datasets, and applications.
Shares AI, SaaS
Chinese multinational technology company specializing in internet services and artificial intelligence.
Shares AI
Aurora Innovation develops the Aurora Driver, a self-driving system for autonomous trucks and ride-hailing vehicles.
Shares AI
Key facts
- Headquarters
- San Francisco, United States
- Total funding
- $150M (Series B)
- Business model
- B2B
Frequently asked
What does Goodfire do?
The platform has two components: one for the LLM training phase and another for monitoring models in production. They claim to have reduced AI hallucinations by half in a recent project. Goodfire developed a method called SPD to understand LLM data processing by identifying and removing components one by one to see their impact on output.
How much funding has Goodfire raised?
Goodfire has raised a total of $150M in funding. The most recent round on record is Series B.
SEC filings, insider activity, nonprofit links and pay benchmarks for Goodfire:
Company details: from public sources. Funding: no announced round on file. Ownership: not yet claimed by the company. Scores: measured by named tools; each links to its method. Our rules
