Prime Highlight
- Inception secured $50 million in seed funding to accelerate the development of its diffusion-based AI models for text and code generation.
- Led by Stanford professor Stefano Ermon, the company aims to make AI systems faster, more efficient, and scalable through diffusion-based technology.
Key Facts
- The funding round was led by Menlo Ventures with participation from Microsoft’s M12, Nvidia’s NVentures, Snowflake Ventures, Databricks Investment, Mayfield, and Innovation Endeavors, along with angel investors Andrew Ng and Andrej Karpathy.
- Inception’s Mercury model, already integrated into ProxyAI, Buildglare, and Kilo Code, delivers over 1,000 tokens per second using diffusion-based methods.
Background
Inception, an artificial intelligence startup that creates diffusion-based AI models, raised $50 million in seed funding to boost its research in text and code generation. The funding round was led by Menlo Ventures and included participation from Mayfield, Innovation Endeavors, Microsoft’s M12, Snowflake Ventures, Databricks Investment, and Nvidia’s NVentures. Prominent AI figures Andrew Ng and Andrej Karpathy also joined as angel investors.
Stanford professor Stefano Ermon leads the company. He is a top researcher in diffusion models, a type of AI that improves data step by step instead of creating content word by word. Unlike auto-regressive models like GPT-5 or Gemini, diffusion models handle data as a whole, making processing faster and more efficient.
With the funding, Inception launched a new Mercury model designed for software development. Mercury has already been integrated into platforms such as ProxyAI, Buildglare, and Kilo Code. Ermon said the company’s approach helps reduce both latency and compute cost, making it a promising alternative to existing large language models.
“These diffusion-based LLMs are much faster and more efficient than what everyone else is building,” Ermon said. He added that diffusion models can process multiple operations at once, allowing for over 1,000 tokens per second, compared to the sequential limits of current AI systems.
By using this method, Inception plans to change the entire AI-powered code-generating and text applications landscape, making future AI development faster and easier to scale.
