Inception, a startup focused on diffusion-based artificial intelligence, has raised $50 million in seed funding to accelerate its work in applying the technology to text and software development. The funding round was led by Menlo Ventures, with participation from Mayfield, Innovation Endeavors, Microsoft’s M12 fund, Snowflake Ventures, Databricks Investment, and Nvidia’s NVentures. AI leaders Andrew Ng and Andrej Karpathy also invested.
The company is led by Stanford professor Stefano Ermon, a long-time contributor to diffusion model research. While diffusion models are best known for image generation tools such as Stable Diffusion and Midjourney, Inception aims to extend the approach to broader applications. The company released a new version of its Mercury model, tailored for coding tasks and already integrated into tools including ProxyAI, Buildglare, and Kilo Code.
Ermon says diffusion models offer advantages over autoregressive systems used in most large language models. By processing information in parallel rather than sequentially, they can improve latency and reduce compute demand. Inception claims its models can achieve speeds above 1,000 tokens per second, highlighting the potential efficiency gains for large-scale development workloads.
